<?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 Ferdynand on Medium]]></title>
        <description><![CDATA[Stories by Ferdynand on Medium]]></description>
        <link>https://medium.com/@ferdynandchan?source=rss-64614c3334b0------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*vRJ8OdjSAH0IsYuVFveoXA.png</url>
            <title>Stories by Ferdynand on Medium</title>
            <link>https://medium.com/@ferdynandchan?source=rss-64614c3334b0------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 19:21:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ferdynandchan/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[Forget the ERP: I Built a Dynamic, Self-Cleaning Asset Tracker with Google Sheets and Saved…]]></title>
            <link>https://ferdynandchan.medium.com/forget-the-erp-i-built-a-dynamic-self-cleaning-asset-tracker-with-google-sheets-and-saved-1911cf45e594?source=rss-64614c3334b0------2</link>
            <guid isPermaLink="false">https://medium.com/p/1911cf45e594</guid>
            <dc:creator><![CDATA[Ferdynand]]></dc:creator>
            <pubDate>Mon, 03 Nov 2025 08:29:18 GMT</pubDate>
            <atom:updated>2025-11-05T05:07:47.162Z</atom:updated>
            <content:encoded><![CDATA[<h4><strong>Forget the ERP: I Built a Dynamic, Self-Cleaning Asset Tracker with Google Sheets and Saved Ourselves Countless Headaches</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Dzw1_DoAyWdApjmE" /><figcaption>Photo by <a href="https://unsplash.com/@rubaitulazad?utm_source=medium&amp;utm_medium=referral">Rubaitul Azad</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>If you’ve ever managed physical assets, projectors, laptops, tools, or even vehicles you know the pain of the inventory spreadsheet. It starts clean, but quickly descends into a Spreadsheet Graveyard.</p><ul><li><em>John logs “projector A-1” as </em><strong><em>A-1</em></strong><em>.</em></li><li><em>Sarah logs the same projector later as </em><strong><em>A1</em></strong><em>.</em></li><li><em>Someone else misspells the location or forgets to update the status when they return the item.</em></li></ul><p>The result? Your “master” inventory sheet is a mess of duplicates and outdated information. You spend more time cleaning data than actually managing assets. We needed a system that wasn’t just a ledger, but a <strong>self-governing data engine</strong>. That’s why I created <strong>DAPTRACK: The Dynamic Automated Projector Tracker.</strong></p><h3>What is DAPTRACK? A System Designed for Consistency</h3><p>DAPTRACK is a robust asset tracking solution built entirely on <strong>Google Sheets and Google Forms</strong>. It’s designed specifically to combat the three biggest enemies of asset management: <strong>inconsistent data, lack of real-time status, and manual upkeep.</strong></p><p>Unlike complex Enterprise Resource Planning (ERP) software, DAPTRACK is simple, accessible, and runs itself once you set it up. Its purpose is twofold: <strong>keep perfect records</strong> and <strong>train future operators</strong> on asset management best practices.</p><h3>The 4 Layers of DAPTRACK Management</h3><p>The system is powered by a simple yet brilliant architecture that separates data entry from data summary and analysis:</p><ol><li><strong>The Google Form (Data Entry):</strong> This is the user’s only interface. It forces consistent entry by validating the <strong>Asset Tag ID</strong> (it <strong>must</strong> include a dash, e.g., Test-01), ensuring that every log is clean. Crucially, if a user submits an invalid format (like Test01), the form throws an error message: <em>&quot;Invalid format! The asset tag must include a dash, like Test-01.&quot;</em> This immediately enforces data standards.</li><li><strong>The </strong><strong>Logs Sheet (The Immutable Record):</strong> Every form submission becomes a single line here. This sheet records the <strong>timestamp, asset tag, user, status, and location</strong>. It is the absolute historical record of all asset activities. <strong>No one touches this data directly.</strong></li><li><strong>The </strong><strong>Grade List Sheet (The Readiness Scorecard):</strong> This acts as the <strong>Predefined support sheet for grading</strong>. It links the unique <strong>Asset ID</strong> (e.g., HT-08) to a fixed, long-term performance score (e.g., <strong>Grade A</strong>).</li><li><strong>The </strong><strong>Status Sheet (The Real-Time Dashboard):</strong> This is the ultimate dashboard. Using powerful <strong>LOOKUP and </strong><strong>ARRAYFORMULA</strong> functions, it automatically scans the <em>entire</em> Logs sheet to pull the <strong>LATEST</strong> status entry for every unique asset, then cross-references it with the <strong>Grade List</strong> to show the asset’s score.</li></ol><h3>The Magic Under the Hood: Self-Cleaning Rules</h3><p><strong>DAPTRACK</strong> is more than just centralization. It’s <strong>smart automation</strong> that eliminates manual decision-making and data cleaning.</p><p>Stop overpaying for data quality. DAPTRACK gives you <strong>enterprise-level consistency</strong> by prioritizing structure at the point of entry and automation at the point of summary.</p><p>Ready for a smarter asset management tool that can be custom-fit for <strong>any asset</strong>, not just projectors?</p><p><a href="https://ferdynandporto.wordpress.com/">📧<strong>Click here to contact me for a demo and purchasing information</strong></a></p><p><strong>Update:</strong><br>So turns out <strong>DAPTRACK</strong> is well received. So, on my own initiative, I went all out in my free time to upgrade the existing DAPTRACK into a newer version of it. Introducing <strong>DASTRACK!</strong></p><p>DASTRACK is basically just DAPTRACK, but now I’m able to track other asset types instead of just projectors. The first entry other than projectors is currently a <strong>toolkit</strong>. Yeah, it might not seem like much, but wait till you see what’s inside the toolkit and count how many items need to be tagged (it’s 12). But this will serve as a humble beginning for my latest project.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1911cf45e594" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Visualization of using Correlation Analysis to Guide Nickel Exploration in Southeast Sulawesi]]></title>
            <link>https://ferdynandchan.medium.com/visualization-of-using-correlation-analysis-to-guide-nickel-exploration-in-southeast-sulawesi-fb79e74aa88f?source=rss-64614c3334b0------2</link>
            <guid isPermaLink="false">https://medium.com/p/fb79e74aa88f</guid>
            <category><![CDATA[prospecting]]></category>
            <category><![CDATA[visualization]]></category>
            <category><![CDATA[lab-sample]]></category>
            <category><![CDATA[mining]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Ferdynand]]></dc:creator>
            <pubDate>Thu, 24 Oct 2024 09:32:04 GMT</pubDate>
            <atom:updated>2024-10-24T09:32:04.544Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/350/0*S9HXSuN-iJ54q1HO" /></figure><h3>Small overview of the data source</h3><p>In Southeast Sulawesi, the abundance of natural resources is immense, particularly the wealth of nickel found in mountain rocks. The government has granted several companies permission to conduct mining activities to extract nickel.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*d1cp0ZCaHVgN9J3j" /></figure><p>The data in this graph comes from lab results obtained after our company sent soil or ore samples for analysis. The analysis provides the concentration of the most commonly found minerals in the mining area. This information helps determine whether it is viable to excavate for nickel in that region.</p><h3>Findings for Ni Above 1.45</h3><p>In this analysis, the top ten correlations for Ni above 1.45 show that the variable most positively correlated with Ni is Al2O3, with a correlation value of 0.271009. This indicates a moderate relationship between Ni and Al2O3. Other variables that exhibit positive correlations include CaO (0.248565) and SiO2 (0.118903), although with lower values.</p><p>On the other hand, there are several variables that have negative correlations with Ni. The highest negative correlation is observed with Fe (-0.0491748) and MgO (-0.0364865), suggesting a minimal impact of these two variables on Ni.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/402/1*bhFv2BQDbT7uDPQgAVuO1g.png" /></figure><h3>Findings for Ni Below 1.45</h3><p>When the Ni content is below 1.45, the analysis reveals that SiO2 has the highest positive correlation with Ni (0.375778), indicating a stronger relationship compared to the above 1.45 condition. This is followed by MgO (0.314628), suggesting that increases in these two variables may be related to an increase in Ni content.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/406/1*St-W6b3Dgnbna1-b09bzag.png" /></figure><p>However, there are significant negative correlations with variables such as CaO (-0.266589) and Cr2O3 (-0.259707), indicating that when Ni levels are lower, increases in these variables tend to be associated with decreases in Ni content.</p><h3>Conclusion</h3><p>Ni levels above 1.45 tend to have a moderate positive relationship with Al2O3 and CaO, while levels below 1.45 show a stronger positive correlation with SiO2 and MgO. These results can help in understanding the interactions of elements in the context of mineral processing and related applications.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fb79e74aa88f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Discovering What Readers Really Think through Sentiment Analysis of Amazon Book Reviews with…]]></title>
            <link>https://ferdynandchan.medium.com/discovering-what-readers-really-think-through-sentiment-analysis-of-amazon-book-reviews-with-dd1e9ac66392?source=rss-64614c3334b0------2</link>
            <guid isPermaLink="false">https://medium.com/p/dd1e9ac66392</guid>
            <category><![CDATA[distilbert]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[transformers]]></category>
            <category><![CDATA[nlp]]></category>
            <dc:creator><![CDATA[Ferdynand]]></dc:creator>
            <pubDate>Thu, 24 Oct 2024 09:29:44 GMT</pubDate>
            <atom:updated>2024-10-24T09:29:44.281Z</atom:updated>
            <content:encoded><![CDATA[<h3>Discovering What Readers Really Think through Sentiment Analysis of Amazon Book Reviews with DistilBERT</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*tD6PuVW6cXCVbuFs" /></figure><h3>Data source and overview</h3><p>The dataset used is published by Mohamed Bekheet under the title <em>Amazon Books Reviews</em>. It contains reviews written by readers and buyers about various books, including text reviews, ratings, and more. This dataset served as a foundation to help the model learn aspect-based sentiment analysis (ABSA), enabling it to identify specific aspects of the books being reviewed and determine the sentiment expressed towards those aspects.</p><p><a href="https://www.kaggle.com/datasets/mohamedbakhet/amazon-books-reviews">https://www.kaggle.com/datasets/mohamedbakhet/amazon-books-reviews</a></p><h3>About the model</h3><p>This model employs Aspect-Based Sentiment Analysis (ABSA) to enhance the book review experience on platforms like Amazon. By analyzing reviews, it goes beyond simple positive or negative ratings to extract specific sentiments about key aspects such as the plot, characters, and writing style. For example, it can reveal insights like “readers loved the plot but found the ending disappointing,” helping potential buyers grasp the strengths and weaknesses that affect a book’s overall rating. This detailed analysis not only aids consumers in making informed purchasing decisions but also provides valuable feedback for authors and publishers to improve their writing and marketing strategies.</p><h3>ABSA MODEL</h3><p>This model trained a DistilBERT on 3 million Amazon book reviews for aspect-based sentiment analysis (ABSA), specifically targeting likes and dislikes. Ratings were categorized into negative, neutral, and positive, with key aspects like “plot” and “characters” identified in the reviews. The model achieved an impressive accuracy of 94–99% in recognizing sentiments, though it encountered challenges with neutral reviews.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dd1e9ac66392" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Visualizing the Key Factors and Sales Drivers in Walmart’s Weekly Revenue and What Influences it]]></title>
            <link>https://ferdynandchan.medium.com/unveiling-the-key-factors-and-sales-drivers-in-walmarts-weekly-revenue-and-what-influences-it-60996693dd25?source=rss-64614c3334b0------2</link>
            <guid isPermaLink="false">https://medium.com/p/60996693dd25</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[walmart]]></category>
            <category><![CDATA[data-visualization]]></category>
            <category><![CDATA[insights]]></category>
            <category><![CDATA[sales]]></category>
            <dc:creator><![CDATA[Ferdynand]]></dc:creator>
            <pubDate>Thu, 24 Oct 2024 09:26:57 GMT</pubDate>
            <atom:updated>2024-10-24T09:41:15.065Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/600/0*d5YgI-jPJOvBAFQ_.png" /></figure><p><strong>Introduction</strong></p><p>Walmart, the retail behemoth with a global presence, operates a vast network of stores offering a wide array of products at everyday low prices. To maintain its competitive edge in a dynamic economic environment, understanding the factors influencing its sales performance is crucial. This research investigates several key economic variables and their impact on Walmart’s weekly revenue.</p><p><strong>Purpose of Research</strong></p><p>The purpose of this research is to understand the key factors influencing Walmart’s sales performance. By analyzing variables such as holidays, temperature, fuel prices, the Consumer Price Index (CPI), and unemployment rates, <strong>This analysis is aimed to uncover insights that could inform business strategies and marketing decisions</strong>. This analysis not only reveals consumer behavior trends but also highlights how economic conditions can impact retail sales which also <strong>provides valuable information for retailers looking to optimize their operations.</strong> Personally, <strong>my motivation</strong> stems from a desire to enhance my knowledge in data analysis and contribute valuable insights to the retail industry.</p><p><strong>About the Dataset</strong></p><p>The “Walmart Sales” dataset on Kaggle provides valuable insights into the factors influencing revenue at Walmart stores across the United States spanning from the year 2010 to 2012. It includes historical sales data for 45 Walmart stores, alongside crucial external factors such as holidays, temperature, fuel prices, CPI, and unemployment rates. This dataset allows for an analysis of the impact of these variables on Walmart’s sales performance and facilitates an understanding of broader economic trends. It can be used to investigate how external economic factors and holiday seasons affect retail sales.</p><p><strong>THE FACTORS</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ZkXih-K7id9txiqA" /></figure><p>While holidays are generally associated with increased consumer spending, our analysis reveals a surprising trend. Average weekly sales during holiday weeks are USD 2,807,220, compared to USD 2,885,148 during non-holiday weeks, indicating a 3% decrease. This suggests that consumers may be diverting their spending towards holiday promotions offered by competitors or adjusting their purchasing behavior due to changes in disposable income.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*NkRQ3NX-qtLxXMU7" /></figure><p>Weather patterns, particularly temperature, subtly influence consumer behavior and, consequently, Walmart’s sales. Our findings demonstrate a correlation between cooler temperatures (below 53.8°F) and a 6.23% increase in average weekly sales. This could be attributed to factors like seasonal shopping events, increased demand for winter-related products, a shift towards indoor activities leading to home improvement projects, and potential psychological factors influencing consumer behavior.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*CzuJBVztWvHsjcge" /></figure><p>Fuel prices, a significant external factor, can impact consumer spending. However, our analysis reveals a minimal impact of fuel price fluctuations on Walmart’s sales. This resilience can be attributed to several factors, including Walmart’s diverse product range, emphasis on essential groceries, commitment to “Everyday Low Prices,” strategic promotions, and a robust supply chain that helps mitigate the impact of fuel price increases on consumer prices.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*XXsIh-dyqR5InXI5" /></figure><p>Inflation, measured by the Consumer Price Index (CPI), can significantly influence consumer spending. Our research indicates an inverse relationship between CPI and Walmart’s sales. As overall prices rise, consumer spending at Walmart tends to decrease, suggesting that inflation can negatively impact sales volume.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*m9DZeYaf3tk4qYbD" /></figure><p>Unemployment rates serve as a critical indicator of economic stability and consumer confidence. Our analysis reveals a strong correlation between rising unemployment and decreased sales at Walmart. When unemployment is high, consumer spending power and confidence are negatively affected, leading to reduced sales.</p><p><strong>Conclusion</strong></p><p>The analysis reveals intriguing and potentially counter-intuitive insights into the factors driving Walmart’s weekly sales. While conventional wisdom might suggest increased sales during holiday weeks, our findings indicate a potential decrease, warranting further investigation into consumer behavior and competitive dynamics within the retail landscape. Conversely, the positive correlation between cooler temperatures and sales presents a valuable opportunity for Walmart to optimize its seasonal product assortment and promotional strategies.</p><p>One particularly noteworthy finding is Walmart’s resilience to fuel price fluctuations, a testament to the effectiveness of its “Everyday Low Prices” strategy and sophisticated supply chain management. This capability to absorb external cost pressures without significantly impacting consumer prices reinforces Walmart’s competitive advantage. However, the sensitivity of sales to CPI and unemployment underscores the undeniable influence of macroeconomic factors on consumer spending, highlighting the need for proactive strategies to mitigate potential downturns.</p><p>For more details, you can see the implementation and further analysis in this <strong>Google Colab </strong>link below:</p><p><a href="https://colab.research.google.com/drive/1V3tinfoPeSAe447St7dVT7Ib07bU2CDc?usp=sharing">https://colab.research.google.com/drive/1V3tinfoPeSAe447St7dVT7Ib07bU2CDc?usp=sharing</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=60996693dd25" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Meningkatkan Sistem Pengambilan Informasi dengan Model Information Retrieval]]></title>
            <link>https://ferdynandchan.medium.com/meningkatkan-sistem-pengambilan-informasi-dengan-model-information-retrieval-ab20a355962d?source=rss-64614c3334b0------2</link>
            <guid isPermaLink="false">https://medium.com/p/ab20a355962d</guid>
            <dc:creator><![CDATA[Ferdynand]]></dc:creator>
            <pubDate>Tue, 09 May 2023 10:30:24 GMT</pubDate>
            <atom:updated>2023-05-09T10:30:24.896Z</atom:updated>
            <content:encoded><![CDATA[<p>Oleh : Angga, Mong, dan Bianka</p><h3>Pendahuluan</h3><p>Dalam era informasi saat ini, akses cepat dan efektif terhadap sumber daya ilmiah menjadi sangat penting. Untuk mencapai tujuan ini, sistem pengambilan informasi yang efisien diperlukan. Dalam penelitian ini, kami mengembangkan sebuah Sistem Information Retrieval dengan menggunakan Vector Space Model dan dataset Vaswani dari ir_datasets. Kami akan menjelaskan teknik-teknik seperti TF-IDF dan cosine similarity yang digunakan dalam implementasi kami, serta membandingkan kinerja model yang dibuat dengan model referensi yang dikembangkan oleh NileyGF.</p><h3>Mengapa Memilih Model Information Retrieval dan Dataset Vaswani?</h3><h3>Dataset Vaswani</h3><p>Salah satu alasan memilih dataset Vaswani adalah ketersediaannya yang luas dalam penelitian Information Retrieval dan Pemrosesan Bahasa Alami. Meskipun diketahui bahwa dataset ini memiliki beberapa kelemahan, kami memutuskan untuk tetap menggunakan dataset ini karena kepopulerannya dan familiaritas yang ada dalam komunitas penelitian. Hal ini memungkinkan kami untuk membandingkan hasil eksperimen kami dengan penelitian lain yang menggunakan dataset yang sama.</p><h3>Tantangan yang Ditawarkan oleh Dataset Vaswani</h3><p>Dataset Vaswani sering kali disebut sebagai dataset yang “mengacaukan eksperimen orang.” Hal ini dikarenakan dataset ini memiliki rata-rata presisi dan recall yang rendah dibandingkan dengan koleksi lainnya. Hal ini menunjukkan bahwa sistem temu kembali informasi yang menggunakan dataset ini seringkali gagal menemukan dokumen yang relevan atau menghasilkan terlalu banyak dokumen yang tidak relevan. Namun, kami melihat tantangan ini sebagai peluang untuk mengembangkan metode dan pendekatan baru yang dapat meningkatkan kinerja model. Dengan menggunakan dataset yang kontroversial seperti Vaswani, kami berharap dapat menemukan wawasan baru dan strategi inovatif dalam pemrosesan informasi.</p><h3>Model NileyGF sebagai Baseline</h3><p>Model NileyGF yang dikembangkan oleh [nama pengembang] digunakan sebagai model referensi dalam perbandingan kinerja. Kami memilih model ini karena memiliki implementasi yang solid dan memadai. Dengan membandingkan model yang kami kembangkan dengan model NileyGF, kami dapat menilai kemajuan yang telah kami capai dan menunjukkan keunggulan dari metode yang kami usulkan.</p><h3>Metode dan Pendekatan yang Digunakan</h3><p>Dalam penelitian ini, kami menggunakan teknik-teknik text mining seperti Term Frequency-Inverse Document Frequency (TF-IDF) dan cosine similarity untuk mengukur relevansi antara permintaan pencarian dan dokumen.</p><p>TF-IDF adalah metode pembobotan yang menghitung bobot suatu kata dalam dokumen berdasarkan kemunculan kata tersebut dalam dokumen itu sendiri dan kumpulan dokumen yang lebih besar. Dalam TF-IDF, setiap dokumen direpresentasikan sebagai vektor yang menunjukkan bobot kata dalam dokumen tersebut. Bobot ini kemudian digunakan untuk menghitung skor relevansi antara permintaan pencarian dan dokumen dengan menggunakan metode cosine similarity.</p><p>Cosine similarity adalah metode yang digunakan untuk mengukur kesamaan antara dua vektor dalam ruang vektor. Dalam konteks pencarian informasi, cosine similarity digunakan untuk mengukur kesamaan antara vektor permintaan pencarian dan vektor dokumen. Semakin tinggi nilai cosine similarity, semakin relevan dokumen dengan permintaan pencarian.</p><h3>Perbandingan Model dengan NileyGF</h3><p>Dalam perbandingan kinerja model yang kami kembangkan dengan model NileyGF, kami fokus pada skor presisi (precision) dan skor recall. Precision mengukur seberapa akurat sistem dalam mengambil dokumen yang relevan, sedangkan recall mengukur seberapa banyak dokumen relevan yang berhasil diambil oleh sistem.</p><p>Berdasarkan hasil pengujian, kami melihat bahwa model yang kami kembangkan menghasilkan rata-rata skor presisi yang lebih tinggi (0.2) dibandingkan dengan model NileyGF (0.1319) dari total 93 query yang diuji. Hal ini menunjukkan bahwa model kami cenderung menghasilkan lebih banyak query yang relevan secara akurat.</p><p>Selain itu, dalam hal skor recall, model yang kami kembangkan juga menunjukkan kinerja yang lebih baik dengan rata-rata skor recall sebesar 0.9 dibandingkan dengan model NileyGF (0.2195). Model kami mampu mengambil kembali sebagian besar, jika tidak semua, query yang relevan dari total query relevan yang ada.</p><h3>Kesimpulan</h3><p>Berdasarkan perbandingan kinerja, model yang kami kembangkan menunjukkan kinerja yang lebih baik dibandingkan dengan model NileyGF. Dengan rata-rata skor presisi yang lebih tinggi dan rata-rata skor recall yang lebih tinggi, kami percaya bahwa model kami mampu menghasilkan hasil pencarian yang lebih relevan dan akurat.</p><p>Meskipun dataset Vaswani memiliki kelemahan dan dapat mengacaukan eksperimen, kami melihat tantangan ini sebagai kesempatan untuk mengembangkan metode dan pendekatan baru. Dengan melakukan penelitian yang komprehensif dan melibatkan diri dalam tantangan yang ada, kami berharap dapat memberikan kontribusi yang bermanfaat dalam pengembangan bidang information retrieval.</p><p>Selengkapnya, Anda dapat melihat implementasi model dan analisis lebih lanjut di repositori kami:</p><p><a href="https://colab.research.google.com/drive/1AizUfUtzFvk0tgKrrBAT8iKhZ5oD8RzH?usp=sharing">Google Colaboratory</a></p><p>Terima kasih telah membaca artikel ini.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ab20a355962d" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>