<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Chang Xiao on Medium]]></title>
        <description><![CDATA[Stories by Chang Xiao on Medium]]></description>
        <link>https://medium.com/@chang-63990?source=rss-ac0a0572bdf------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*cyxCw_hee3i89SkIo3zzEA.png</url>
            <title>Stories by Chang Xiao on Medium</title>
            <link>https://medium.com/@chang-63990?source=rss-ac0a0572bdf------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 19:58:32 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@chang-63990/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Zwift Race Recommendation using simple machine learning]]></title>
            <link>https://chang-63990.medium.com/zwift-race-recommendation-using-simple-machine-learning-ff1091ff7461?source=rss-ac0a0572bdf------2</link>
            <guid isPermaLink="false">https://medium.com/p/ff1091ff7461</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[zwift]]></category>
            <category><![CDATA[recommendation-system]]></category>
            <category><![CDATA[cycling]]></category>
            <dc:creator><![CDATA[Chang Xiao]]></dc:creator>
            <pubDate>Mon, 31 Oct 2022 18:42:19 GMT</pubDate>
            <atom:updated>2022-10-31T18:42:19.773Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*BOBBcfAxMtQwj3Yf" /><figcaption>Photo by <a href="https://unsplash.com/@markusspiske?utm_source=medium&amp;utm_medium=referral">Markus Spiske</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Zwift racing is one of the coolest aspects of the virtual cycling platform. It has been growing both from a grassroots level as well as officially (UCI e-sports worlds, etc). However, it is also one of the most heated discussion areas as many complaints about the current watts-per-kilogram (WKG) category system.</p><p>The idea of using machine learning to enhance the racing experience has been attempted before. For example, the <a href="https://brunogregory.medium.com/predicting-winners-in-cycling-races-with-machine-learning-b3d7f1126513">ZRace app</a> tries to predict a racer’s placement in a race given that racer’s profile (power numbers, HR, etc).</p><p>However, I believe the key to applying machine learning to Zwift is more of an unsupervised learning problem, specifically the question:</p><blockquote>How do we match the races to a user that optimizes his/her race experience</blockquote><p>In other words, how do we <em>recommend races</em> to users that they will enjoy?</p><h3><strong>Our Hypothesis</strong></h3><p>We assume that a rider would “enjoy” a race if they have placed high on the result in whatever category they raced in. We then can assign a “rating” based on their result in their perspective categories. For example:</p><ul><li>Top 3 (podium) — rating = 5</li><li>Top 10% (outside of top 3) — rating = 4</li><li>Top 25% — rating = 3</li><li>Top 50% — rating = 2</li><li>50% and below — rating = 1</li></ul><p>The ranking percentile is important since some races may have 10 riders and some may have more than 200.</p><h3>Our Machine Learning Pipeline</h3><p>1. We collected my own racing data from Zwiftpower.com and generated implicit ratings for the races I participated in.</p><p>2. We also collected historical racing data from other racers that participated in the same races as me and generated implicit ratings based on their performance.</p><p>3. The dataset is fed into a KNN model with a grid search for the best model parameters using the <a href="https://surpriselib.com/">Python Surprise library</a>.</p><p>4. We built a method to generate the top rating predictions for a user to provide recommendations for the event/race courses they might do well.</p><p>5. We built a method to find “similar riders” to check if they are reasonable for potential “matchmaking”</p><h3>Testing the model</h3><p>For the model testing, we looked at traditional metrics for model accuracies such as RSME, and MAE. However, we also checked individual users for a sanity test to see if the recommendations made sense.</p><h4><strong>Rating prediction</strong></h4><p>We first tested what kind of events the algorithm will predict a user will like e.g. the user would rate these events very high (4 to 5)</p><p>I used my racing profile as the sample to test.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xfHpdrSEYIuy6VpGpT4X8Q.png" /><figcaption>Race profile on Zwiftpower</figcaption></figure><p>I have raced on Zwift 37 times according to Zwiftpower, I have a decent racing record with a number of podium finishes. I have raced on all types of courses: Flat e.g. Crit City, London Classique; Punchy: e.g. Innsbruck ring, Perimeter loop (New York) and Mountainous: e.g. Mountain Route, Road to Sky.</p><p>I would summarize myself as a punchy rider who does well on routes with sub-5-minute climbs. My best race course is probably the Innsbruck ring.</p><p>Based on the implicit ratings from my racing records, I have an average rating of <strong>3.67</strong>. This means any race prediction <strong>above this rating</strong> can be recommended to me.</p><p><strong>The top races/courses recommended to me were:</strong></p><ul><li>Crit City Race — The Bell lap (7 occurrences)</li><li>3R Watopia Flat Sprint Race — 1 Lap (7 occurrences)</li><li>Crit City Race — Downtown Dolphin (6 occurrences)</li><li>3R Racing — Innsbruck Ring ( 6 occurrences)</li><li>Various Races (sprint races to multi-lap races) — London Classique (5 occurrences)</li></ul><p>There were other top predictions such as race courses that included <em>Road to Sky, Keith Hill after party, and Tempus Fugit Time Trial.</em></p><p>The recommendation was both good and somewhat surprising. I have only raced Crit City — Downtown Dolphin once (albeit a winning one). The most surprising was that <em>Tempus Fugit Time Trial</em> was recommended 3 times but I have never done a TT race on this course before since I am a lightweight rider who doesn’t do well on flat time trials.</p><p>The initial test shows the system does capture my racing preference and recommends to me races/courses that I can do well at.</p><h4><strong>Finding Similar Riders</strong></h4><p>We can also use the algorithm to find similar riders from the dataset using the KNN algorithm, this has many interesting applications including race matchmaking.</p><p>Again I used my profile to test finding “similar riders”, the result was interesting as well.</p><p><strong>Similar rider 1<br></strong>The first rider who matched with me had done two of the same races as me. His result in the 2 races was below mine.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*B70_9y3GNjan28Nmb0J1Dw.png" /><figcaption>Rider 1</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*p7CAWvr4w6B8AKSCroX_Wg.png" /><figcaption>My Record</figcaption></figure><p><strong>Similar rider 2<br></strong>The second rider was someone I remember from the races we did together. Our results at the races were very close.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KEeyl6HaX-X0fKbWJ4av7Q.png" /><figcaption>Rider 2</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IJqos7UqzxEVtG56qkg-QQ.png" /><figcaption>My Record</figcaption></figure><p>Other similar riders were less interesting as the similarity score decreased. This is expected from my limited racing history (37) as well as the overall dataset we had for testing.</p><h3>Initial Observations</h3><p>There were some interesting takeaways from the initial testing:</p><ul><li>Race recommendations (predictions) were positive, I was recommended for race courses that I would do well at as well as others I wouldn’t think about racing often.</li><li>Finding similar riders did not perform as well as I had expected, this is mostly due to the lack of racing history from me as well as the limited dataset.</li><li>It’s possible to build a recommendation system with heuristic rules such as common races participated by riders, and race courses where riders have placed well. However, it will run into the risk of a closed feedback loop (echo chamber).</li></ul><h3>Application of the recommendation system</h3><p><strong>Recommend users to follow</strong><br>Off the bat, this can be used to suggest other riders follow on Zwift. These will be riders that performed well on similar races as you.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*dyOOjKDr-sC9g0_k1R7rLw.png" /><figcaption>User follow screen</figcaption></figure><p><strong>Recommend similar races<br></strong>We can use our rating prediction for a user and find upcoming races/courses that are similar to the past races the user has done well at.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*JxLQAvYctmy1smS4YBre6w.png" /><figcaption>Race recommendation</figcaption></figure><h3>About me</h3><p>I am an avid cyclist who enjoys riding outside or on Zwift. I am also a technologist and data science hobbyist. I have been interested in improving the racing experience for fellow Zwifters using Machine Learning techniques.</p><p>The purpose of the blog post is purely for research and personal interest.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ff1091ff7461" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How many types of product recommendations do you need? (Shein vs. Amazon approach)]]></title>
            <link>https://medium.com/siggy-recommender/how-many-types-of-product-recommendations-do-you-need-shein-vs-amazon-approach-59d30c551fe9?source=rss-ac0a0572bdf------2</link>
            <guid isPermaLink="false">https://medium.com/p/59d30c551fe9</guid>
            <category><![CDATA[siggy]]></category>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[product-recommendations]]></category>
            <dc:creator><![CDATA[Chang Xiao]]></dc:creator>
            <pubDate>Mon, 18 Jul 2022 12:02:31 GMT</pubDate>
            <atom:updated>2022-07-18T12:02:31.558Z</atom:updated>
            <content:encoded><![CDATA[<p>Product recommendations aka “You recently viewed”, “Related items”, and “What others bought” have become common occurrences nowadays. There are first popularized by Amazon as a tool for upselling and cross-selling products (sales conversion/optimization). Now you can find product recommendations on most major e-commerce stores.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hJeeh1QmWizizIZuxZlepg.png" /><figcaption>(Product Recommendations From Amazon.com)</figcaption></figure><p>The idea of product recommendations is similar to traditional physical stores where products are organized in departments and in extreme cases such as Ikea where the store layout is designed intentionally to keep you shopping longer.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EmeEkiDBNsNkMs4zqWWdlA.png" /><figcaption>Ikea Store Layout</figcaption></figure><h4>The many types of product Recommendations</h4><p>Some of the common product recommendations include:</p><p><strong><em>Frequently Bought Together</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YyPvJN3PCYOW3ZkiEkNk4A.png" /><figcaption>Frequently Bought Together</figcaption></figure><p>The basic psychology behind this type of recommendation is to upsell additional complementary products the shopper might need. (For example, show bicycle lock, spare tube, and helmet when someone is looking at purchasing a bike)</p><blockquote>Tip: look for a recommendation app that allows you to offer discount or additional incentive when purchasing bundled products</blockquote><p><strong><em>What Others Liked/Bought/Viewed</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*L6HLNL7lqHvsBF04SbMyPA.png" /><figcaption>Example from Walmart.com</figcaption></figure><p>This type of product recommendation looks at shoppers browsing behavior and presents items that may be interesting to users that are deemed “similar”. It is one of the <a href="https://www.amazon.science/the-history-of-amazons-recommendation-algorithm">first popularized machine-learning product recommendation algorithms from Amazon.</a></p><blockquote>Tip: This is good for recommending items across different product categories since shoppers browsing session can vary dramatically. The products recommended also refreshes to reflect the seasonality (e.g. Christmas items during the holidays)</blockquote><p><strong><em>Similar Products</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XmHJi3cYLrH_gAcQ5u3s_g.png" /><figcaption>Related Products from Amazon</figcaption></figure><p>This type of product recommendation will present items that are similar (e.g. Show bigger or smaller TVs when a shopper is looking at a TV). The primary benefit of this is to minimize abandonment if a shopper does not find what they are looking for immediately.</p><blockquote>This type of recommendation can be machine-learning based or manual such as tagging/taxonomy-based</blockquote><p><strong><em>Other types…….</em></strong></p><p>There are also other types of product recommendations such as “recently viewed items”, “popular items”, “products on sale”, etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/861/1*AnCSWUxopxaJ7D9SzZZhKg.png" /><figcaption>Recently Viewed</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ekyk8vjxyRjA29o42KtPvg.png" /><figcaption>Popular items</figcaption></figure><h4>One Recommendation (rules them all?)</h4><p>Instead of presenting many different product recommendations on the screen, some stores only present one set of product recommendations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/452/1*LWaOlSzBjbGuBEI-b2UsIA.gif" /></figure><p>A primary example of this is Shein.com, one of the most popular e-commerce stores at the moment. Shein presents one set of product recommendations that continues to load, offering users a full browsing experience.</p><p>The recommendations are a mix of similar products and what the user has recently viewed.</p><h4>The bottom line</h4><p>Product recommendations have proved they are here to stay. However, there is no definite answer to your approach in presenting multiple recommendations (Amazon approach) or a single recommendation type (Shein approach).</p><p>To know what works for you still requires extensive A/B testing and data analytics.</p><p><a href="https://medium.com/siggy-recommender/how-can-a-product-recommendation-system-help-your-e-commerce-store-b076e09748fd">How can a product recommendation system help your e-commerce store?</a></p><p><em>Siggy is a Machine-learning/AI-based product recommender for Shopify. If you run a Shopify Store, you can </em><a href="http://www.siggy.ai/"><em>Try Siggy Free</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=59d30c551fe9" width="1" height="1" alt=""><hr><p><a href="https://medium.com/siggy-recommender/how-many-types-of-product-recommendations-do-you-need-shein-vs-amazon-approach-59d30c551fe9">How many types of product recommendations do you need? (Shein vs. Amazon approach)</a> was originally published in <a href="https://medium.com/siggy-recommender">Siggy Recommender</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How can a product recommendation system help your e-commerce store?]]></title>
            <link>https://medium.com/siggy-recommender/how-can-a-product-recommendation-system-help-your-e-commerce-store-b076e09748fd?source=rss-ac0a0572bdf------2</link>
            <guid isPermaLink="false">https://medium.com/p/b076e09748fd</guid>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[product-recommendations]]></category>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[siggy]]></category>
            <dc:creator><![CDATA[Chang Xiao]]></dc:creator>
            <pubDate>Mon, 11 Jul 2022 10:02:46 GMT</pubDate>
            <atom:updated>2022-07-11T10:02:46.597Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*8BJ0DBrgu5Dg2QsX" /><figcaption>Photo by <a href="https://unsplash.com/@markkoenig?utm_source=medium&amp;utm_medium=referral">Mark König</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3>(Why) Do you need a product recommendation system?</h3><p>Product recommendation systems have been around for a while. They have become more relevant in recent years with the rise of Amazon and machine-learning/AI technologies.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fDMykfcnIlmuWa92EdZ58g.png" /><figcaption>Customers also bought…</figcaption></figure><p>For example, Amazon’s famous “<em>Customers also bought</em>” product recommendation.</p><p>Below are some general reasons why you could use a product recommendation system if you are not Amazon.</p><h4>Reason #1: If you have a large product catalog</h4><p>A product recommendation system is useful when your shop has a large number of products. There is no general rule to this, however, if your store has at least 100 products it can be suitable.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nTU8u0ZT13PszORYYCcccQ.png" /><figcaption>Example Product Catalog with 500+ products</figcaption></figure><p>This mostly plays into the “<em>comparable shopper” </em>psychology. The product recommendation can help capture a potential sale if the shopper sees a similar or comparable item in your shop. This can also help reduce abandonment as the recommendation can help the shopper to find alternative products from your shop.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rfEs436FHHfwJHFER_z0SQ.png" /><figcaption>Similar Products</figcaption></figure><h4>Reason #2: If your store has high/sufficient traffic volume</h4><p>The product recommendation system sits in the bottom half of the conversion funnel. This means it is only helpful if your shop already receives sufficient traffic from both organic and paid channels.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/778/1*1QFpLrIxQD1cnXRkpAGC9w.png" /><figcaption>Example Conversion Funnel for Product Recommendation</figcaption></figure><h4>Reason #3: If you are looking to bundle products</h4><p>If you sell products that are complementary to each other, product recommendations can help make additional sales.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/881/1*yRPg-BaRIYNd-p3Vu6O1rQ.png" /><figcaption>(Bundle example from Zara)</figcaption></figure><p>The classic example of this includes the “Shop this look” or “Wear with” where an apparel store can recommend items from a specific outfit set to the shopper.</p><h4>Other reasons</h4><p>There are other reasons to use a product recommendation system including promoting discounted/overstocked items, sponsored/affiliated product listings, etc.</p><h3>How can product recommendations help your bottom line?</h3><p>The best way to test whether a product recommendation system can help your shop is through:</p><ol><li>Identify the Key Performance Metrics(KPM)</li><li>Analytics tracking of the KPMs</li><li>Extensive A/B testing</li></ol><p><strong>Identify the key performance metrics (KPM)</strong></p><p>For example, some common KPM to track for e-commerce include <em>average order value, average basket size, cart abandonment rate, and others</em>.</p><p><strong>Analytics tracking of the KPMs</strong></p><p>The above KPMs should be tracked through analytics software such as Google Analytics or as part of your e-commerce shop platform. Additionally, the recommendation engine should offer conversion analytics such as click rate of recommended items, added to cart, and checkout statistics.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/643/0*qi-jOTjE7VaQ4OCv.png" /><figcaption>Example Analytics Tracking in Siggy.ai Product Recommender</figcaption></figure><p><strong>A/B Testing</strong></p><p>To truly know if a recommendation system is working or not, extensive A/B testing and multivariate testing are required. For example, the simplest A/B testing will be to split your site traffic so that half of the shoppers will see product recommendations and the other half will not.</p><p>The resulting KPMs from the two sets of audiences can give you insight into if the recommendation system is working well for your e-commerce shop.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Z3w4IN-plW8ntl7KJGidFQ.png" /><figcaption>Example A/B Testing</figcaption></figure><h3>Common types of product recommendation systems</h3><h4>Curated Recommendations</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*YQYq3sqVyoX7ftVv" /><figcaption>Photo by <a href="https://unsplash.com/@mtimber71?utm_source=medium&amp;utm_medium=referral">Mark Timberlake</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>This type of recommendation system allows you to <strong>manually add products</strong> to be recommended to the visitor. For example, adding accessories related to an outfit.</p><h4>Taxonomy-based (Tagging) Recommendations</h4><p>This type of recommendation system relies on the tags entered for the product and it displays other products that have the same or similar tags.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/943/1*x59l3Dsczg1xMvItqdRG8Q.png" /><figcaption>(Shopify Product Taxonomy)</figcaption></figure><h4>Machine-learning/AI-based (Data-driven) Recommendations</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Y0MjKJStOUUZgV_r" /><figcaption>Photo by <a href="https://unsplash.com/@deepmind?utm_source=medium&amp;utm_medium=referral">DeepMind</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>This type of recommendation system uses various data sources to present recommendations to shoppers. The most common data sources include product catalog data, user visits, product ratings, reviews, order information, and others.</p><p><em>Siggy is a Machine-learning/AI-based product recommender for Shopify. If you run a Shopify Store, you can </em><a href="http://www.siggy.ai/"><em>Try Siggy Free</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b076e09748fd" width="1" height="1" alt=""><hr><p><a href="https://medium.com/siggy-recommender/how-can-a-product-recommendation-system-help-your-e-commerce-store-b076e09748fd">How can a product recommendation system help your e-commerce store?</a> was originally published in <a href="https://medium.com/siggy-recommender">Siggy Recommender</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Customize Siggy Recommendation Widget placement]]></title>
            <link>https://medium.com/siggy-recommender/customize-siggy-recommendation-widget-placement-6d846f42cae1?source=rss-ac0a0572bdf------2</link>
            <guid isPermaLink="false">https://medium.com/p/6d846f42cae1</guid>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[siggy]]></category>
            <category><![CDATA[product-recommendations]]></category>
            <dc:creator><![CDATA[Chang Xiao]]></dc:creator>
            <pubDate>Mon, 20 Jun 2022 19:04:13 GMT</pubDate>
            <atom:updated>2022-06-20T19:04:13.380Z</atom:updated>
            <content:encoded><![CDATA[<p>Siggy tries to place its product recommendations below the main product area by default.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*evuRNAGq4WetJPt9r4SQrA.png" /><figcaption>Default Recommendation placement</figcaption></figure><p>However, sometimes you may wish to place the widget in other areas on the product page due to various reasons (e.g. Siggy incorrectly placed the recommendation widget, placing the widget above the product reviews section, etc).</p><p>For example, if I wanted to place the recommendations below the <strong>“BUY IT NOW” </strong>button.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dIhPpxEk4GO1dAjUyMJFKw.png" /></figure><h4>1. Go to your current theme in your Shopify Store</h4><p>You can get there from “Sales channels” -&gt; “Online Store” -&gt; “Themes”. You will see the active theme.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/989/1*GdTwNxaIpXxx75evnkom4A.png" /><figcaption>Online Store -&gt; Themes</figcaption></figure><p>Click on “Actions” -&gt; Edit code</p><h4>2. Find the product-template.liquid theme file</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/549/1*OHYPNbwaxwi2_yjP2hQ2OA.png" /><figcaption>product-template.liquid</figcaption></figure><h4>3. Locate where you would like to place the widget</h4><p>In our case, I would like to place the recommendations below the “<strong>BUY IT NOW</strong>” button</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1010/1*Ee57KEP-xyjTF3NfAFzqlg.png" /><figcaption>product-template.liquid</figcaption></figure><h4>4. Add the Siggy HTML tag</h4><p>Add &lt;div id=&quot;siggy-app&quot;&gt;&lt;/div&gt; in the desired area of the template and save the template.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/932/1*sCitwmLuLzxtcciXLHhi6A.png" /></figure><p>The recommendation widget will now be placed at the new location on the page.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*PJvP69Be61sbUq-mElnipg.png" /><figcaption>Custom placement</figcaption></figure><p>If you are interested, install <a href="http://www.siggy.ai/">Siggy</a> for free for Shopify!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6d846f42cae1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/siggy-recommender/customize-siggy-recommendation-widget-placement-6d846f42cae1">Customize Siggy Recommendation Widget placement</a> was originally published in <a href="https://medium.com/siggy-recommender">Siggy Recommender</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Siggy.ai + Infinite Recommendation = The Future]]></title>
            <link>https://medium.com/siggy-recommender/siggy-ai-infinite-recommendation-the-future-372b189d3209?source=rss-ac0a0572bdf------2</link>
            <guid isPermaLink="false">https://medium.com/p/372b189d3209</guid>
            <category><![CDATA[siggy]]></category>
            <category><![CDATA[product-recommendations]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[shopify]]></category>
            <dc:creator><![CDATA[Chang Xiao]]></dc:creator>
            <pubDate>Fri, 13 May 2022 14:54:09 GMT</pubDate>
            <atom:updated>2022-05-13T14:54:09.100Z</atom:updated>
            <content:encoded><![CDATA[<p>It’s rare that we are excited about a single feature. However, we firmly believe this feature is a game-changer for Siggy.ai and content recommendations.</p><h3>Standard Recommendation</h3><p>Traditionally, we only display a set number of product recommendations (e.g.) to the user ranked as the most relevant based on the combination of our algorithms.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/390/1*yEsL-X9aJbuqoNJ7lumhRg.png" /><figcaption>(Traditional recommendations)</figcaption></figure><p>The shortfall of the standard recommendation is that the user often views this as a secondary type of content that they usually don’t pay a lot of attention to.</p><h3>“Infinite Recommendation ”</h3><blockquote>“Infinite recommendation” promotes product recommendation as part of the user’s content browsing experience.</blockquote><p>In the infinite recommendation mode, we continue to show recommendations generated by our algorithm as the user scroll through the page. This changes the user experience similar to browsing a collection or a catalog in an online store.</p><p><a href="https://www.youtube.com/shorts/7qeGZFibbR0">Siggy Infinite Recommendation</a></p><h3>Turn on the “Infinite Recommendation”</h3><p>You can turn on the “infinite recommendation” mode under the “widget customizer” settings in siggy:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ivbVaNKhwTGPVBG1krPafQ.png" /></figure><p>What are you waiting for? <a href="https://apps.shopify.com/siggy-product-recommender">Get Siggy for Free for your Shopify Store</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=372b189d3209" width="1" height="1" alt=""><hr><p><a href="https://medium.com/siggy-recommender/siggy-ai-infinite-recommendation-the-future-372b189d3209">Siggy.ai + Infinite Recommendation = The Future</a> was originally published in <a href="https://medium.com/siggy-recommender">Siggy Recommender</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[5 Simple tips for data modeling with Cassandra/Datastax Astra]]></title>
            <link>https://chang-63990.medium.com/5-simple-tips-for-data-modeling-with-cassandra-datastax-astra-42a53300846c?source=rss-ac0a0572bdf------2</link>
            <guid isPermaLink="false">https://medium.com/p/42a53300846c</guid>
            <category><![CDATA[datastax]]></category>
            <category><![CDATA[nosql]]></category>
            <category><![CDATA[data-modeling]]></category>
            <category><![CDATA[cassandra]]></category>
            <dc:creator><![CDATA[Chang Xiao]]></dc:creator>
            <pubDate>Wed, 16 Mar 2022 13:50:59 GMT</pubDate>
            <atom:updated>2022-03-16T13:50:59.218Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*vTmtrYe76F0mKXs0" /><figcaption>Photo by <a href="https://unsplash.com/@kaip?utm_source=medium&amp;utm_medium=referral">Kai Pilger</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>It is easy to apply relational database data modeling concepts to Cassandra. However, you may quickly find it to be counter-productive for your application. After some painful data modeling refactoring for our application <a href="http://www.siggy.ai/">Siggy.ai</a>, here are some quick tips:</p><h4>1. The partition key does not have to be unique</h4><p>Although the simplest design is a partition key that is the same as the primary key. This means you will have 1 row per partition. You can often assign a partition key that encompasses more than 1 row of data. For example, some partition keys can be a tag, date, or even a bucket value (e.g. 1, 2, 3)</p><h4><strong>2. Design partition key to controlling your data size</strong></h4><p>The partition key defines where the data is stored and how it is replicated. The rule of thumb is less than 100MB or 100,000 rows of data per partition.</p><h4>3. Add columns to the primary key only for the uniqueness</h4><p>The primary key can contain the partition key. When you have a partition that contains more than 1 row it may be good to add more fields to query for unique records. For example: in a blog table you may use a date field as the partition key. Adding blog post title field and author field to the primary key will most likely give you the ability to query for a unique blog post by date, blog post title, and the author.</p><h4>4. Data ordering with clustering columns</h4><p>The non-partition key fields (mentioned above) in your primary key are also called clustering columns. They are very useful for your inequality query (&gt;, &lt;, etc) as well as how your data is ordered by default.</p><p>For example: if we had a blog table with a partition key of a “date”, we can add a clustering column called “hour (0–23)”. You can now query blog posts created on a specific date and within a specific range of hours of that day.</p><p>Additionally, you can also define the default ordering of data with the clustering key (ascending or descending order). This means if you query for the first 10 blog posts from a day, it will be ordered by the hour of that day based on the ordering definition.</p><h4>5. TimeUUID will be your best friend</h4><p>Lastly, <a href="https://docs.datastax.com/en/cql-oss/3.3/cql/cql_reference/timeuuid_functions_r.html">TimeUUID</a> can often be used to form the primary key that provides both <strong>uniqueness</strong> and <strong>sorting </strong>capabilities. You can convert timeUUID into date/time values as well as perform range queries with timeUUIDs.</p><h4>Useful resources</h4><p>If you are interested to dive deeper into the best practices in data modeling. Datastax has a free data modeling course that can be very helpful for you to avoid any pitfalls when transitioning to Cassandra.</p><p><a href="https://academy.datastax.com/resources/ds220-data-modeling">https://academy.datastax.com/resources/ds220-data-modeling</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=42a53300846c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Out-of-stock product recommendations: explained]]></title>
            <link>https://medium.com/siggy-recommender/out-of-stock-product-recommendations-explained-79474bd716c7?source=rss-ac0a0572bdf------2</link>
            <guid isPermaLink="false">https://medium.com/p/79474bd716c7</guid>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[product-inventory]]></category>
            <dc:creator><![CDATA[Chang Xiao]]></dc:creator>
            <pubDate>Wed, 09 Mar 2022 13:03:28 GMT</pubDate>
            <atom:updated>2022-03-09T13:03:28.060Z</atom:updated>
            <content:encoded><![CDATA[<p>One of the most requested features for Siggy is the ability to display recommendations for <strong>out-of-stock </strong>products. Although Siggy supports this by default, we will dive into the Shopify configurations a bit further to explain how this works.</p><h4>The one to rule them all!</h4><p>If you just want to enable out-of-stock items to be displayed from the product recommendation, you can simply check the “<em>Display out of stock products</em>” configuration in Siggy.</p><p><em>(Please note this will ignore your Shopify inventory tracking setting for the product and may display products that are not meant to be shown)</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*izHMMabs-cExaR2mw9JhTQ.gif" /><figcaption>Out of stock configuration</figcaption></figure><p>To better understand how Siggy will provide recommendations to out-of-stock items by default, read on the following example setup in Shopify.</p><h4><strong>Case 1: Inventory not tracked (Shown in recommendation)</strong></h4><p>Siggy will <em>display the product in the recommendation</em> if inventory is not tracked for all variants of a product (e.g. selling digital products, using a different system for tracking inventory).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zTRtQ82LeUf5ubMyB31mmw.png" /></figure><h4>Case 2: Inventory tracked and Out-of-stock (Not shown in recommendation)</h4><p>Siggy will <em>not display the product in the recommendation</em> if inventory is tracked for one or more variants and the variant has 0 quantity (out of stock).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QQj0sFj3qouj1bGmhDam3Q.png" /></figure><h4>Case 3: Selling out of stock items</h4><p>Siggy will <em>display the product in the recommendation</em> if inventory is tracked for one or more variants, the variant is out of stock and the option “<strong>Continue selling when out of stock” </strong>is checked</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JTRdx50iCX0qN6WauQ_f1Q.png" /></figure><h3>About Siggy</h3><p><a href="http://www.siggy.ai/">Siggy.ai</a> is an AI-based product recommendation app for Shopify. You can <a href="https://apps.shopify.com/siggy-product-recommender">install Siggy for free</a> in Shopify.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=79474bd716c7" width="1" height="1" alt=""><hr><p><a href="https://medium.com/siggy-recommender/out-of-stock-product-recommendations-explained-79474bd716c7">Out-of-stock product recommendations: explained</a> was originally published in <a href="https://medium.com/siggy-recommender">Siggy Recommender</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Recommending on sale products in Siggy]]></title>
            <link>https://medium.com/siggy-recommender/recommending-on-sale-products-in-siggy-9408a7878ed3?source=rss-ac0a0572bdf------2</link>
            <guid isPermaLink="false">https://medium.com/p/9408a7878ed3</guid>
            <category><![CDATA[siggy]]></category>
            <category><![CDATA[shopify]]></category>
            <category><![CDATA[product-recommendations]]></category>
            <dc:creator><![CDATA[Chang Xiao]]></dc:creator>
            <pubDate>Wed, 09 Mar 2022 13:03:08 GMT</pubDate>
            <atom:updated>2022-03-09T13:03:08.344Z</atom:updated>
            <content:encoded><![CDATA[<p>It is common for Shop owners to put items on sale after the holiday season to clear out inventory. Siggy can help this by prioritizing on-sale items in its recommendation.</p><p>For example, below is the default recommendations generated by Siggy:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q3p1k_aJMSfjHKKR2dxDfg.png" /><figcaption>Default recommendation</figcaption></figure><p>Notice that two of the items recommended are on sale. However, the top two recommendations are not.</p><figure><img alt="Siggy configuration" src="https://cdn-images-1.medium.com/max/1024/1*PGiLGbD1pMX-ujkNpN3d5Q.gif" /><figcaption>Siggy configuration</figcaption></figure><p>You can go into Siggy’s configuration and turn on the “<strong>Display on-sale items first</strong>”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*F4c11TkuA6E6E42lns97sg.png" /><figcaption>New Recommendations</figcaption></figure><p>You will note that the new recommendation list has shifted up the two on-sale items to the front of the list and pulled in additional on-sale items.</p><h3>About Siggy</h3><p><a href="http://www.siggy.ai/">Siggy.ai</a> is an AI-based product recommendation app for Shopify. You can <a href="https://apps.shopify.com/siggy-product-recommender">install Siggy for free</a> in Shopify.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9408a7878ed3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/siggy-recommender/recommending-on-sale-products-in-siggy-9408a7878ed3">Recommending on sale products in Siggy</a> was originally published in <a href="https://medium.com/siggy-recommender">Siggy Recommender</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Product recommendation with implicit ratings and anonymous users.]]></title>
            <link>https://medium.com/siggy-recommender/product-recommendation-with-implicit-ratings-and-anonymous-users-1fa42b6784fd?source=rss-ac0a0572bdf------2</link>
            <guid isPermaLink="false">https://medium.com/p/1fa42b6784fd</guid>
            <category><![CDATA[knn]]></category>
            <category><![CDATA[product-recommendations]]></category>
            <category><![CDATA[siggy]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[collaborative-filtering]]></category>
            <dc:creator><![CDATA[Chang Xiao]]></dc:creator>
            <pubDate>Wed, 22 Dec 2021 20:32:00 GMT</pubDate>
            <atom:updated>2021-12-22T20:33:34.715Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/984/1*EmsRuVnBzT1D2_uWup3mgw.png" /><figcaption>(Product recommendation)</figcaption></figure><p>The classic collaborative filtering algorithm is useful when you have a group of <strong>specific users</strong> that <strong>explicitly provide ratings</strong> to many items. This is challenging for many online stores because:</p><ol><li>A user may shop anonymously without creating a user account.</li><li>Users do not provide ratings or reviews of products.</li></ol><p>This post will discuss how you can apply the collaborative filtering algorithm under these constraints. If you are interested, read on!</p><h3>Our machine learning pipeline</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EQVH9BvQaR9PWeE0HbIVWA.png" /><figcaption>(The machine learning pipeline)</figcaption></figure><p>We have a very simple overview that shows the key parts of the machine learning pipeline.</p><ol><li><strong>Data processing — </strong>This should be run frequently to stream and process the user visit data and store them as rating data.</li><li><strong>Machine learning — </strong>This should be run periodically (e.g. once a week) to (re)train the data.</li><li><strong>App/Presentation — </strong>The frontend application will use the KNN generated from the last trained algorithm.</li></ol><h3>Generate implicit ratings</h3><p>When users are not providing explicit ratings to products in a online shop, we will have to derive this from user behaviors. Some of the key user behaviors can include:</p><ul><li><strong>Time spent on a product page — </strong>Longer time spent infers the user is interested in the product by reading the product information, etc.</li><li><strong>Secondary actions — </strong>Some actions such as adding the product to a shopping wishlist or watching the product when it’s out of stock can also signify users’ interest level.</li><li><strong>Primary actions — </strong>Actions such as adding the product to the cart or purchasing the product (checkout) can be very good indicators of the user’s interest level in the product.</li></ul><p>The next step is to create some rules based on selected user behaviors and generate an implicit rating for the user/product.</p><p><em>For example, a very simple rule can be (For a rating scale of 1 to 5):</em></p><p>If the user spends more than 2 minutes on a product page, we give them a rating of 5; between 1 and 2 minutes a rating of 4, less than 1 minute a rating of 3 (to denote an average/neutral rating).</p><p>The time spent should be based on the overall distribution of time spent on product pages (e.g. 2 minutes is 2 standard deviations from the mean, etc).</p><p><strong>You can also create complex rules to generate implicit ratings based on a combination of user behaviors. The best approach is to test different rules and measure the effectiveness of the recommendation.</strong></p><blockquote>For technical implementation, anonymous user behavior data can be obtained from analytics software such as <em>Google Analytics, Adobe Analytics, Matomo, </em>etc</blockquote><h3>Process ratings</h3><p>Depending on the amount of traffic your online store receives, you can stream the visitor traffic and generate ratings in real-time or fetch them periodically (e.g. every 30 minutes).</p><p>If your store is seasonal (e.g. Holiday products, the Spring collection, etc), you can also consider pruning the rating data periodically. This can potentially make the recommendation more dynamic and reflect the seasonality of the user behavior.</p><blockquote>It is only worth considering data pruning in this example since we will not be predicting the top product recommendations for a user. In that case, more data is helpful to improve the accuracy of the algorithm.</blockquote><h3>Algorithm and Data training</h3><p>At this point, we should have a rating table consisting of at least the <em>user id, product id, rating (1–5)</em>.</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/d55428d187ef3a4132661a81285cfcf6/href">https://medium.com/media/d55428d187ef3a4132661a81285cfcf6/href</a></iframe><blockquote>Sometimes it is helpful to keep additional data columns (such as human-understandable product names) to quickly experiment and do sanity checks on the recommendations.</blockquote><p>To train the data, we can use the python <a href="https://github.com/NicolasHug/Surprise">Surprise library</a>. It is a very simple, easy-to-use library with many collaborative-filtering algorithms.</p><p>Let’s look a simple example:</p><pre>from surprise import Dataset, Reader, KNNWithMeans</pre><pre>reader = Reader(line_format=&#39;user item rating&#39;, sep=&#39;,&#39;, rating_scale=(1,5))<br>data = Dataset.load_from_file(&#39;/path/to/ratings.csv&#39;, reader=reader)<br>trainingSet = data.build_full_trainset()</pre><pre>sim_options = {<br>  &quot;name&quot;: &quot;cosine&quot;, <br>  &quot;min_support&quot;: 3,<br>  &quot;user_based&quot;: False<br>}</pre><pre>algo = KNNWithMeans(sim_options = args)<br>algo.fit(trainingSet)</pre><p>Here we are essentially doing the following:</p><ol><li>Load our rating data from a CSV file into the training set. In a production environment, you can also load the data from a Pandas dataframe.</li><li>We build our <em>TrainingSet </em>using the entire dataset. We do not need to split the dataset into train/test since we do not need to run prediction and measure our algorithm’s accuracy.</li><li>We specify a list of reasonable options for our algorithm. <br>For example, use <em>cosine</em> distance (instead of <em>msd</em>) to measure the similarity between the products. <br>The min_support is the number of common items between users before we consider them for similarity.<br>user_based is <em>False </em>to denote we will be using item-based collaborative filtering.</li></ol><blockquote>We chose item-based collaborative filtering because a user will likely only rate a few items. Since we view each visitor as a unique user (anonymous), we will end up with a lot of users that rate a small part of the product catalog. On the flip side, an item can be rated by a lot of users. The item-based approach means working with a more densely populated matrix.</blockquote><blockquote>Generally speaking, item-based collaborative filtering works better when you have more users than items.</blockquote><p>Here we use a <a href="https://surprise.readthedocs.io/en/stable/knn_inspired.html#surprise.prediction_algorithms.knns.KNNWithMeans">KNNwithMeans </a>algorithm. It is a memory-based collaborative filtering algorithm that works fast. SVD and other model-based algorithm are more accurate if you are running predictions.</p><h3>Find KNN and present recommendations</h3><p>After training the data, the algorithm will calculate a <em>similarity matrix</em>. This will be a matrix with a shape of (n,n) where n = the number of items (products) in the trainingset. It will contain the <em>cosine distance </em>of all products to all other products. (The matrix should also have a diagonal of zeros since the product’s distance to itself will be 0)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/933/1*wrQwkfDDTI-5BK-cocmz5A.png" /><figcaption>Example of the similarity matrix subsection with the diagonal of zeros (0)</figcaption></figure><blockquote>On some implementations, the diagonal of the similarity matrix will be ones (1), that’s using the similarity factor (1- the cosine distance)</blockquote><p>Our goal will be “<em>given a product, find its KNN (K Nearest Neighbors).</em>” Another way to say this is give me the products that are most similar to a given product (based on user ratings).</p><pre>k = 4 # We are looking for the top 4 similar products<br>sim_list = [] <br>knn = [] # contain the product ids of the k similar products </pre><pre>product_id = &#39;123456&#39;<br>innerId = trainingSet.to_inner_iid(product_id)</pre><pre># The exclusion id sequence, if the KNN is 0 to 4 means we are not getting a valid KNN<br>ex_id_seq = {0:0,1:1,2:2,3:3,4:4}<br>if innerId in ex_id_seq.keys():<br>  del ex_id_seq[innerId]</pre><pre># Finding the k nearest neighbors using the similarity matrix<br>neighbors = algo.get_neighbors(innerId, k)<br>for iid in neighbors:<br>  sim_list.append(iid)</pre><pre># Check to see if the sims list contains the exclusion id sequence<br>check_default_ids = set(ex_id_seq.keys()).issubset(sim_list)</pre><pre>if check_default_ids:<br>  # Find the position (index) within the sim_list<br>  first_ex_id = list(ex_id_seq.values())[0]<br>  sim_list = sim_list[:sim_list.index(first_ex_id)]</pre><pre>for id in sim_list:<br>  knn.append(trainingSet.to_raw_iid(id))</pre><p>There are a few quirks in finding the KNN:</p><p>The algorithm stores the <em>user_ids</em> and <em>product_ids</em> as <em>innerIDs</em> so you will always need to perform a conversion such as</p><pre>trainingSet.to_inner_iid(product_id)</pre><p>When the algorithm fails to find the k nearest neighbors for a product, it will always return the first k items from the dataset (0 to k-1). We look for this sequence and make sure to not include it in our list of recommendations.</p><p><strong>Finally, an example recommendation using this algorithm looks like this:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cwvNU1F6ae1JvjcimCXY0g.png" /><figcaption>(Example recommendation)</figcaption></figure><h3>Other observations</h3><p>The algorithm is highly dependent on having an adequate amount of traffic to the online store, therefore it will not be effective if the online store receives little traffic (e.g. new stores).</p><p>The algorithm will likely only provide recommendations to a subset of the product catalog. This means less popular products or new products that receive very few ratings will not show up on the recommendation.</p><p>It’s always better to build a recommendation system using a combination of algorithms. For example, using product data and content filtering can help complement the recommendations from collaborative filtering alone.</p><h3>Next steps</h3><p>The approach we have taken so far focused on the recommendation scenario of <em>“given a product, show me N related products based on user’s implicit ratings”</em>.</p><p>We can better take advantage of collaborative filtering algorithms by making predictions of the top product recommendation for a given user. This means “<em>given a user, show me the top N products recommended for this user.</em>”</p><p>To accomplish this, we will need to:</p><ul><li>Capture and generate live ratings for a user currently browsing the online store.</li><li>Find the most similar user based on the implicit ratings of the current user from the previously trained algorithm.</li><li>Make the top N prediction for that “most similar user” and present it to the current user.</li></ul><p>In this scenario, it is worth experimenting with different machine learning algorithms such as SVD and others to optimize the prediction accuracy.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1fa42b6784fd" width="1" height="1" alt=""><hr><p><a href="https://medium.com/siggy-recommender/product-recommendation-with-implicit-ratings-and-anonymous-users-1fa42b6784fd">Product recommendation with implicit ratings and anonymous users.</a> was originally published in <a href="https://medium.com/siggy-recommender">Siggy Recommender</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Display product price in Siggy product recommendation widget]]></title>
            <link>https://medium.com/siggy-recommender/display-product-price-in-siggy-product-recommendation-widget-b0c35f5a9cb3?source=rss-ac0a0572bdf------2</link>
            <guid isPermaLink="false">https://medium.com/p/b0c35f5a9cb3</guid>
            <category><![CDATA[product-recommendations]]></category>
            <category><![CDATA[siggy]]></category>
            <category><![CDATA[shopify]]></category>
            <dc:creator><![CDATA[Chang Xiao]]></dc:creator>
            <pubDate>Thu, 09 Dec 2021 21:43:59 GMT</pubDate>
            <atom:updated>2021-12-09T21:43:59.558Z</atom:updated>
            <content:encoded><![CDATA[<p>You can now perform additional customization to the recommendation widget with the price display toggle.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DHdzYjAEMt5fOM8DnvoidA.png" /><figcaption>(Recommendation widget with price displayed)</figcaption></figure><p>To enable the price display, see the video below:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F-Nnl0FbisAs%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3D-Nnl0FbisAs&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F-Nnl0FbisAs%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/729653cec8de5729a8bf7e63ec80f453/href">https://medium.com/media/729653cec8de5729a8bf7e63ec80f453/href</a></iframe><h3>Multi-currency support</h3><p>Our widget supports shops that accept multiple currencies. However, the currency format may be displayed differently depending on how you set your primary store language.</p><p>For example, if your store’s primary language is configured as “English” and the user selects Danish Krone (DKK) as their currency. It will be displayed as “<em>DKK 258.05</em>”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/901/1*Iwa8xMkwwaMDSvxZH6mAmA.png" /><figcaption>(Primary language)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*h55FyQxaxQfRCEETe9fIig.png" /><figcaption>(DKK display for English language store)</figcaption></figure><p>If the primary store language is configured as <strong>“Danish”</strong>. The price display will appear as 258,05 kr:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fQC93P1HYYfibIp8DJpmRQ.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b0c35f5a9cb3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/siggy-recommender/display-product-price-in-siggy-product-recommendation-widget-b0c35f5a9cb3">Display product price in Siggy product recommendation widget</a> was originally published in <a href="https://medium.com/siggy-recommender">Siggy Recommender</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>