<?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 Koride Nagaraju on Medium]]></title>
        <description><![CDATA[Stories by Koride Nagaraju on Medium]]></description>
        <link>https://medium.com/@koride.nagaraju?source=rss-5530e48852a8------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*MoTPO9UHfXXx87Nq.</url>
            <title>Stories by Koride Nagaraju on Medium</title>
            <link>https://medium.com/@koride.nagaraju?source=rss-5530e48852a8------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 17:10:12 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@koride.nagaraju/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[Demand Forecasting and ML]]></title>
            <link>https://medium.com/analytics-vidhya/demand-forecasting-and-ml-8ee9484f6fd2?source=rss-5530e48852a8------2</link>
            <guid isPermaLink="false">https://medium.com/p/8ee9484f6fd2</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[supply-chain]]></category>
            <category><![CDATA[statistical-modeling]]></category>
            <category><![CDATA[demand-forecasting]]></category>
            <dc:creator><![CDATA[Koride Nagaraju]]></dc:creator>
            <pubDate>Tue, 13 Oct 2020 12:56:23 GMT</pubDate>
            <atom:updated>2020-10-13T12:56:23.380Z</atom:updated>
            <content:encoded><![CDATA[<p>Demand Forecasting getting almost right is the dream come true for any supply chain executive.</p><p>It obviously drives all your challenges from shortages of supply , inventory excess to customer satisfaction.</p><p>In this article, we discuss how and where one can leverage AI in demand forecasting.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*cSL_SmpMTEG7kt3c" /><figcaption>Photo by <a href="https://unsplash.com/@evan__bray?utm_source=medium&amp;utm_medium=referral">Evan Dennis</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3>Once upon a time….</h3><p>As per Harvard Business Review, dated 1971, there are essentially three basic types for demand forecast.</p><blockquote>Qualitative techniques, (Expert panels, market research etc)</blockquote><blockquote>Time series analysis and projection (moving average, exponential smoothing etc)</blockquote><blockquote>Causal models. (regression, econometric models etc)</blockquote><p><a href="https://hbr.org/1971/07/how-to-choose-the-right-forecasting-technique">How to Choose the Right Forecasting Technique</a></p><h3>In more recent times….</h3><p>Fast forward to more recent times,</p><p>The prestigious M4 Forecasting competition 2018 gave some surprising insights on impact of ML on Demand forecasting.</p><blockquote>Out Of the 17 most accurate methods, 12 were “combinations” of mostly statistical approaches.</blockquote><blockquote>The biggest surprise was a “hybrid” approach that utilized both statistical and ML features.</blockquote><blockquote>The second most accurate method was a combination of seven statistical methods and one ML one, with the weights for the averaging being calculated by a ML algorithm that was trained to minimize the forecasting error through holdout tests</blockquote><blockquote>The six pure ML methods performed poorly, with none of them being more accurate than the combination benchmark and only one being more accurate than Naïve2.</blockquote><p><em>Source: </em><a href="https://www.sciencedirect.com/science/article/abs/pii/S0169207018300785"><em>https//www.sciencedirect.com/science/article/abs/pii/S016920701830078</em></a></p><p>The results clearly gave an indication that</p><blockquote>Combination of statistical models worked better than a single model</blockquote><blockquote>Hybrid of statistical and machine learning stood a better chance.</blockquote><h3><strong>How do I decide:</strong></h3><ol><li><em>You have done all you could using statistical modelling approach in terms of causality (influence of variables on output), econometry (influence of economic factors over output) , qualitative analysis (expert opinions) but still left with multiple models giving comparable results.</em></li></ol><p>Machine Learning layer can give you that extra edge in deciding the weightage for each model</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*4-Z83UxKX2Y_Dhdp" /><figcaption>Photo by <a href="https://unsplash.com/@karolhere?utm_source=medium&amp;utm_medium=referral">Karol Stefański</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*y-FV4By2xaBru40V" /><figcaption>Photo by <a href="https://unsplash.com/@lunarts?utm_source=medium&amp;utm_medium=referral">Volodymyr Hryshchenko</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>2. <em>Your data variables or features are fairly unintuitive, and you have loads of data. ML techniques can help you in dimensionality reduction.</em></p><p>3. <em>ML may not be your go to technique if your forecast item is a new product with no or little historic data.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ffiaWuopqmFjb1OO" /><figcaption>Photo by <a href="https://unsplash.com/@productschool?utm_source=medium&amp;utm_medium=referral">Product School</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>4. <em>Watch out for paradigm difference between statistical and machine learning models. Statistics is the foundation for both techniques.<br>In statistical modelling, you use up your entire data and come up with a model which “fits” the data.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Pg0CaS4dfCTH4rqb" /><figcaption>Photo by <a href="https://unsplash.com/@raheemsphoto?utm_source=medium&amp;utm_medium=referral">Enayet Raheem</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p><em>In ML, you split data into training, validation and test data sets.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NzwsZlcaZyaefA4B" /><figcaption>Photo by <a href="https://unsplash.com/@herfrenchness?utm_source=medium&amp;utm_medium=referral">Clarisse Croset</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>5. <em>Statistical model gives you much more than a prediction. It can give you confidence interval. Example: There is 95% chance that this year’s demand forecast lies between 200M$ and 250M$.</em></p><p><em>ML techniques in their standard form focus solely on prediction. Example: This year’s forecast is 212M$ with a MAPE of 8% (Mean Absolute Percentage Error), with no confidence interval information.</em></p><blockquote><strong>Thus, when it comes to demand forecasting, statistical and machine learning approaches should not be viewed as contesting techniques, but as collaborative and complementary techniques.</strong></blockquote><p>Hope this article gave you an intuition of impact of ML on Demand Forecast.</p><p>Watch out this space on how to put AI to work for supply chain. Till then,</p><p>Happy Forecasting.</p><p>Bye for now.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*YTvfawEgUNUaH7rA" /><figcaption>Photo by <a href="https://unsplash.com/@iamricozamudio?utm_source=medium&amp;utm_medium=referral">Rico Zamudio</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8ee9484f6fd2" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/demand-forecasting-and-ml-8ee9484f6fd2">Demand Forecasting and ML</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Genesis Of MidAppLife Crisis]]></title>
            <link>https://medium.com/analytics-vidhya/genesis-of-midapplife-crisis-e92756ab0974?source=rss-5530e48852a8------2</link>
            <guid isPermaLink="false">https://medium.com/p/e92756ab0974</guid>
            <category><![CDATA[application-architecture]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[cloud]]></category>
            <dc:creator><![CDATA[Koride Nagaraju]]></dc:creator>
            <pubDate>Sat, 09 May 2020 13:52:49 GMT</pubDate>
            <atom:updated>2020-05-09T13:52:49.581Z</atom:updated>
            <content:encoded><![CDATA[<p>Few years back, you and your team has coded up enterprise application.</p><p>The Application has gradually grown bigger with regular addition of more and more components.</p><p>Fast forward to now, your Application is no longer a baby, it even matured past the teenager like behavioral fluctuations on a whim and is middle aged.</p><h3>In fact unless you act swiftly it may even be facing mid life crisis.</h3><p>The Application is under intense pressure to do the following.</p><ol><li>Be friendly and respectful with old applications (legacy technologies)</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*xHY6qGFIbPYMgwuM" /><figcaption>Photo by <a href="https://unsplash.com/@micheile?utm_source=medium&amp;utm_medium=referral">Micheile Henderson</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>2. Communicate with faster and younger applications in their language. (Ex: You still on JSON? No longer cool. We use Protobuf)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*n8iadLYEw0joEeXJ" /><figcaption>Photo by <a href="https://unsplash.com/@dylandgillis?utm_source=medium&amp;utm_medium=referral">Dylan Gillis</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>3. It cannot cling to its roots. (On-Premise)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*OYDaS101oGVK0NVb" /><figcaption>Photo by <a href="https://unsplash.com/@kate_sade?utm_source=medium&amp;utm_medium=referral">kate.sade</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>4. It does not know which cloud it will go to and for how long it will stay there. (Multi cloud is the new cloud)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*vEHgtcnzfsVnzeB6" /><figcaption>Photo by <a href="https://unsplash.com/@dianamia?utm_source=medium&amp;utm_medium=referral">C Dustin</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>5. It needs to learn to listen more. (Read Callback APIs and feedback mechanisms)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*7A5FDskjEfsIoo0v" /><figcaption>Photo by <a href="https://unsplash.com/@taylor_grote?utm_source=medium&amp;utm_medium=referral">Taylor Grote</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>6. It cant walk too fast (Too many users and transactions are at risk)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*6pHERf5-48teOh8o" /><figcaption>Photo by <a href="https://unsplash.com/@christianfregnan?utm_source=medium&amp;utm_medium=referral">Christian Fregnan</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>7. It cant walk too slow (Remember Agile?)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*gWVWhR6r15p2u0Ys" /><figcaption>Photo by <a href="https://unsplash.com/@areksan?utm_source=medium&amp;utm_medium=referral">Arek Adeoye</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Does this situation resonate with you?</p><p>Hang on for the next article on how to avert and manage “MidAppLife Crisis” leveraging Hybrid Cloud and Artificial Intelligence powered Modernization Framework.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e92756ab0974" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/genesis-of-midapplife-crisis-e92756ab0974">Genesis Of MidAppLife Crisis</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Intuitions of ML Algorithms: Linear Classifiers and Theta Vector Representation]]></title>
            <link>https://medium.com/analytics-vidhya/intuitions-of-ml-algorithms-linear-classifiers-and-theta-vector-representation-ded7ee6c8bcb?source=rss-5530e48852a8------2</link>
            <guid isPermaLink="false">https://medium.com/p/ded7ee6c8bcb</guid>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Koride Nagaraju]]></dc:creator>
            <pubDate>Sun, 01 Mar 2020 15:32:44 GMT</pubDate>
            <atom:updated>2020-03-04T04:19:01.879Z</atom:updated>
            <content:encoded><![CDATA[<p>The intuitions behind theory, representation and deduction of Machine Learning algorithms can serve as great tools for model architecture selection and design in practical applications.</p><p>Lets start off by analyzing a two dimensional data set X with the objective of classifying it.</p><p>For instance, let X represent set of days. Let Each day x in set X is denoted by two features.</p><blockquote>x₁: Temperature at 7 AM</blockquote><blockquote>x₂: Temperature at 4 AM</blockquote><p>Let y represent the classification output whether sun came out on a particular day or not (1 or -1)</p><p>Lets assume, for now, that a linear classifier is possible to achieve above.</p><p><em>( The assumption is for illustration and sun rise event is complex and just two features may not be sufficient to arrive at a linear classification.)</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/801/1*HxxTsINA4X1ual22SeDLAA.jpeg" /></figure><p><strong>Note</strong>: C<strong><em>lassification algorithms, y usually represents the classification output (sunny day (1) or gloomy day (-1)) and not to be confused with coordinate geometry conventions where y usually represents something else.</em></strong></p><p>You can see that if temperature was negative at 4 AM (x2) but if it turned positive at 8 AM (x1) , there is a fairly good chance that day turns out to be sunny.</p><p>So far, so good. From the image above, we do observe that there is a linear classifier decision boundary line in blue which does seems to do a decent job of classifying positive (sunny- +1 ) and negative days (gloomy -1).</p><p>Its called a decision boundary line as classifications on on side of it are positive and the other side are negative.</p><p>How do we represent this line?</p><p>Its passing through the origin, and hence it should be of the form</p><blockquote><strong>A.x1 + B.x2 = 0 : Equation 1</strong></blockquote><p>The above equation has two unknowns A,B. Lets call them</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/219/1*yYz3V2okT1DiSAgKnhT91Q.png" /><figcaption><strong>Introducing theta vector components</strong></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/399/1*283YEuZ_iUIgLuL0P57_PQ.png" /><figcaption>Equation 1 representing decision boundary</figcaption></figure><p>The above equation gives intuition to Equation 2 which is essentially a dot product of input vector [x1,x2] and theta vector.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/925/1*Yi6A4_2kMQwjUc9Fp-dNWw.png" /><figcaption>Equation 2 representing decision boundary</figcaption></figure><p>The dot product is zero and hence theta is a perpendicular to the classifier’s decision boundary. Hence classifier can be described in terms of theta vector.</p><p>The classifier output can be described in this case as</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/471/1*gB15oD4giWzDLHARDSR4eQ.png" /><figcaption>Sunny day classification</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/524/1*mlZd_6j0ZuTQGGJKJiaoBA.png" /><figcaption>Gloomy day classification</figcaption></figure><p>So far, we have seen the intuition for linear classifier and representing decision boundary by a theta vector.</p><p>In the next article, we will look into the intuition behind theory of Decision Margin and Hinge Loss concept for Linear Classifier.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ded7ee6c8bcb" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/intuitions-of-ml-algorithms-linear-classifiers-and-theta-vector-representation-ded7ee6c8bcb">Intuitions of ML Algorithms: Linear Classifiers and Theta Vector Representation</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>