<?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 Nicole Michelle on Medium]]></title>
        <description><![CDATA[Stories by Nicole Michelle on Medium]]></description>
        <link>https://medium.com/@nicolemichelledsouza?source=rss-7fe846709991------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*q6HA_Oz63ovfwOUfAdXbbg.jpeg</url>
            <title>Stories by Nicole Michelle on Medium</title>
            <link>https://medium.com/@nicolemichelledsouza?source=rss-7fe846709991------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 02:30:58 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@nicolemichelledsouza/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[Dealing with Imbalanced Datasets]]></title>
            <link>https://nicolemichelledsouza.medium.com/dealing-with-imbalanced-datasets-4cbfa74af520?source=rss-7fe846709991------2</link>
            <guid isPermaLink="false">https://medium.com/p/4cbfa74af520</guid>
            <category><![CDATA[imbalanced-class]]></category>
            <category><![CDATA[classification]]></category>
            <category><![CDATA[imbalanced-dataset]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[imbalanced-data]]></category>
            <dc:creator><![CDATA[Nicole Michelle]]></dc:creator>
            <pubDate>Fri, 10 Jun 2022 17:41:48 GMT</pubDate>
            <atom:updated>2022-06-10T17:42:59.559Z</atom:updated>
            <content:encoded><![CDATA[<p>Imbalanced data is a very common problem in machine learning, but for a beginner dealing with imbalanced classes can be a little intimidating, since a lot of suggestions on dealing with imbalanced data include stuff like<em> “Collect More Data” </em>and for the novice ML enthusiast trying out their first project on Kaggle collecting more data isn’t an option. This article focuses on what is imbalanced data, why is it a problem and how do you work towards fixing it!</p><p>Imbalanced data is data where the distribution observations across classes is not equal, eg if there are 2 classes and the distribution is 50–50 then that would be called balanced data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/0*_fd8Og5YcpdbYpMb.jpg" /></figure><p>So here’s the thing you will very rarely find data with exactly the same number of rows across classes and while this isn’t that much of a problem when the imbalance is small, if the imbalance is large it can greatly affect the working of the model. for example if the majority class has 900 rows and minority class has 100 rows and the classifier simply labels each row with the majority class it will still claim to have over 90% accuracy, because models are focused on improving accuracy and minimizing error, now to fix this there are 3 simple ways. Please split the train and test dataset before dealing with imbalance and perform these operations <strong>only </strong>on the train set.</p><h3>Over Sampling</h3><p>Over sampling duplicates observations from the minority class until both the minority and majority class are equally distributed, this is great when the minority class is extremely underrepresented, however sometimes oversampling may lead to overfitting as fundamentally we are only duplicating the minority class and the model may try to memorize specific data points.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*CZxPlxNkti7tkBkU.png" /><figcaption><a href="https://dataaspirant.com/handle-imbalanced-data-machine-learning/">https://dataaspirant.com/handle-imbalanced-data-machine-learning/</a></figcaption></figure><h3>Under Sampling</h3><p>Under sampling deletes observations from the majority class until both the minority and majority class are equally distributed, this is works when we have tons of rows of the majority class, however we do this at the cost of valuable information. This could also lead to poor accuracy due to underfitting as we may have deleted important data points.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*BjraHdq4dnwWqBCX.png" /><figcaption><a href="https://dataaspirant.com/wp-content/uploads/2020/08/17-undersampling.png">https://dataaspirant.com/wp-content/uploads/2020/08/17-undersampling.png</a></figcaption></figure><h3>Data Augmentation</h3><p>This can be thought of as a more developed over sampling technique. Using the SMOTE algorithm we can create new data points using the information we have on the existing data points, essentially instead of duplicating data we are generating new data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/656/0*S3nJZ0J5R8x1M3Rd.png" /></figure><p>The way this works is it takes a random sample from the minority class, identifies it’s K nearest neighbors and by moving the data point a little in the direction of the neighbors you generate a synthetic data point. This gives you a data point that isn’t an exact duplicate but isn’t too far away from the actual data point.</p><p>Apart from this you can also look into using a different algorithm for your classification task, preferably a if-else algorithm like decision trees or random forests these are found to be very good with imbalanced data as it addresses both the classes regardless of the distribution.</p><p><em>If you like this article please leave an upvote, if you don’t like it please let me know what I should work on :)</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4cbfa74af520" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to connect YouTube Data API to Android App]]></title>
            <link>https://nicolemichelledsouza.medium.com/how-to-connect-youtube-data-api-to-android-app-81cddfef89ea?source=rss-7fe846709991------2</link>
            <guid isPermaLink="false">https://medium.com/p/81cddfef89ea</guid>
            <category><![CDATA[youtube-api]]></category>
            <category><![CDATA[android-app-development]]></category>
            <dc:creator><![CDATA[Nicole Michelle]]></dc:creator>
            <pubDate>Sat, 19 Mar 2022 03:11:19 GMT</pubDate>
            <atom:updated>2022-03-19T03:11:19.252Z</atom:updated>
            <content:encoded><![CDATA[<p>In this article, I will show you, how to connect the YouTube API to an Android App. Let’s get started!</p><p><strong>Features :</strong></p><p>The YouTube API enables users to</p><ul><li>Retrieve feeds related to videos, users, and playlists.</li><li>It also provides the ability to manipulate these feeds, such as creating new playlists, adding videos as favorites, and sending messages.</li><li>The API is also able to upload videos.</li><li>Register event listeners to get callbacks for certain events, such as the player loading a video or the player state changing.</li><li>Has a helper functionality to support orientation changes as well as transitions to Fullscreen playback.</li></ul><p><strong>Tutorial:</strong></p><p><strong>Prerequisites</strong>: Google Account</p><p>Step 1: Begin with opening Android Studio and Creating a New Project, Select Empty Activity</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*kyLpO-6nnDQ7jpcB" /></figure><p>Step 2: Now before using the YouTube API you need to register your application, for this, we need to do the following</p><ul><li>Go to the Google Developers Console</li><li>Create a New Project(I have named mine SelfLab)</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ywA_KTUOIaxgaZZ6" /></figure><ul><li>Select API Overview</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jUxiMGexbO4N16Hq" /></figure><ul><li>Select Credentials in APIs &amp; Services</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*JXA6gYAx-Re0rg4y" /></figure><ul><li>Select Create Credentials and then select API key, copy the key and we will need it later</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*tCl4eD8TqiSR5ob1" /></figure><ul><li>Select Library and search for YouTube Data API</li><li>Click on Enable API</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*YQViTqzzQqt8daz8" /></figure><ul><li>Now that we have obtained our key and the API, let’s start working on the Android App</li><li>Create a Java Class File called Config.java we will be using this to configure our API, paste the following code in the file.</li></ul><pre>public final class Config {</pre><pre>private Config() {<br> }</pre><pre>public static final String YOUTUBE_API_KEY = “Your API Key Here”;</pre><pre>}</pre><ul><li>Go to the link given below and download the YouTube player API, extract the Jar file and place it in the libs folder</li></ul><p><a href="https://developers.google.com/youtube/android/player/downloads/">YouTube Android Player API — Download | Google Developers</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*BZ0AdX-OSSMuNWVA" /></figure><ul><li>Now in the Gradle folder add the following code in the dependencies</li></ul><pre>compile files(‘libs/YouTubeAndroidPlayerApi.jar’)</pre><ul><li>In the AndroidManifest.xml add the internet permissions</li></ul><pre>&lt;uses-permission android:name=&quot;android.permission.INTERNET&quot;/&gt;</pre><ul><li>Declare the following strings in Strings.xml</li></ul><pre>&lt;resources&gt;<br> &lt;string name=”app_name”&gt;VideoTube&lt;/string&gt;<br> &lt;string name=”player_error”&gt;Error initializing YouTube player: %s&lt;/string&gt;<br> &lt;string name=”seek_to”&gt;Jump To&lt;/string&gt;<br> &lt;string name=”seek_to_hint”&gt;Seconds&lt;/string&gt;<br>&lt;/resources&gt;</pre><ul><li>We will now work towards making the YouTube Player View in the Activity_Main.xml file</li></ul><pre>&lt;RelativeLayout xmlns:android=”http://schemas.android.com/apk/res/android&quot;<br> xmlns:tools=”http://schemas.android.com/tools&quot;<br> android:layout_width=”match_parent”<br> android:layout_height=”match_parent”<br> tools:context=”.MainActivity”&gt;</pre><pre>&lt;com.google.android.youtube.player.YouTubePlayerView<br> android:id=”@+id/youtube_view”<br> android:layout_width=”match_parent”<br> android:layout_height=”wrap_content”/&gt;</pre><pre>&lt;/RelativeLayout&gt;</pre><ul><li>We will now create a class that is a subclass of YouTubeBaseActivity. This is required to make use of YouTubePlayerView. We implemented YouTubePlayer.OnInitializedListener to listen for initialization success or failure. The interface has two methods, named onInitializationFailure() and onInitializationSuccess(). If initialization is successful, the cueVideo() method plays the YouTube video and in case of failure, checks to see whether the error is recoverable by user action.</li><li>If it’s not then a Toast of the error is shown to the user and if it’s user-recoverable, then the getErrorDialog() method shows a dialog that will enable the user to recover from the error.</li></ul><pre>import android.content.Intent;<br>import android.os.Bundle;<br>import android.widget.Toast;</pre><pre>import com.google.android.youtube.player.YouTubeBaseActivity;<br>import com.google.android.youtube.player.YouTubeInitializationResult;<br>import com.google.android.youtube.player.YouTubePlayer;<br>import com.google.android.youtube.player.YouTubePlayer.Provider;<br>import com.google.android.youtube.player.YouTubePlayerView;</pre><pre>public class MainActivity extends YouTubeBaseActivity implements YouTubePlayer.OnInitializedListener {</pre><pre>private static final int RECOVERY_REQUEST = 1;<br> private YouTubePlayerView youTubeView;</pre><pre>@Override<br> protected void onCreate(Bundle savedInstanceState) {<br> super.onCreate(savedInstanceState);<br> setContentView(R.layout.activity_main);</pre><pre>youTubeView = (YouTubePlayerView) findViewById(R.id.youtube_view);<br> youTubeView.initialize(Config.YOUTUBE_API_KEY, this);<br> }</pre><pre>@Override<br> public void onInitializationSuccess(Provider provider, YouTubePlayer player, boolean wasRestored) {<br> if (!wasRestored) {<br> player.cueVideo(“fhWaJi1Hsfo”); // Plays https://www.youtube.com/watch?v=fhWaJi1Hsfo<br> }<br> }</pre><pre>@Override<br> public void onInitializationFailure(Provider provider, YouTubeInitializationResult errorReason) {<br> if (errorReason.isUserRecoverableError()) {<br> errorReason.getErrorDialog(this, RECOVERY_REQUEST).show();<br> } else {<br> String error = String.format(getString(R.string.player_error), errorReason.toString());<br> Toast.makeText(this, error, Toast.LENGTH_LONG).show();<br> }<br> }</pre><pre>@Override<br> protected void onActivityResult(int requestCode, int resultCode, Intent data) {<br> if (requestCode == RECOVERY_REQUEST) {<br> // Retry initialization if user performed a recovery action<br> getYouTubePlayerProvider().initialize(Config.YOUTUBE_API_KEY, this);<br> }<br> }</pre><pre>protected Provider getYouTubePlayerProvider() {<br> return youTubeView;<br> }<br>}</pre><p><em>If you like this article please leave an upvote, if you don’t like it please let me know what I should work on to improve it :)</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=81cddfef89ea" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Machine Learning 101:All things Classification]]></title>
            <link>https://nicolemichelledsouza.medium.com/machine-learning-101-all-things-classification-ee6dff9c2af5?source=rss-7fe846709991------2</link>
            <guid isPermaLink="false">https://medium.com/p/ee6dff9c2af5</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[classification]]></category>
            <category><![CDATA[naive-bayes]]></category>
            <category><![CDATA[ml-so-good]]></category>
            <category><![CDATA[decision-tree]]></category>
            <dc:creator><![CDATA[Nicole Michelle]]></dc:creator>
            <pubDate>Sat, 19 Feb 2022 14:14:27 GMT</pubDate>
            <atom:updated>2022-02-20T10:43:03.488Z</atom:updated>
            <content:encoded><![CDATA[<p>In the last article I spoke about the different types of regression, this time we will dive into Classification and it’s different types</p><p>Classification is a supervised learning algorithm, it is used to predict discrete values like whether you have cancer or not is an email spam or not, which of the 3 economical brackets you belong to, but fundamentally it maps the input data to a specific predefined class like malignant or benign (in tumor detection). As I mentioned in my previous article I will cover Logistic Regression in Classification Algorithms, apart from that we will look at a few others like Decision Tree, KNN and Naïve Bayes</p><h3>Logistic Regression</h3><p>Logistic Regression is similar to Linear Regression except it predicts if something is true or false, instead of fitting a line to the data like in case of Linear Regression, here an S shaped curve called the Sigmoid Function, although logistic regression gives you the probability of true or false it’s mostly used in binary classification problems, Now how this works is if the probability of it being true is greater than 50% it is true else it is false. It’s this ability to provide probabilities and classify samples using continuous and discrete measurements that makes it a popular classification algorithm, apart from binary classification there are many other types of logistic regression like multinomial logistic regression(many classes), ordinal logistic regression (multiple ordered classes).</p><figure><img alt="Logistic Regression: A Simplified Approach Using Python | LaptrinhX" src="https://cdn-images-1.medium.com/proxy/1*zfH9946AssCx4vzjaizWeg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/287/0*6mpc4duWzxOBMAAK" /></figure><p>The sigmoid function is what we use to calculate the probability of something being true or false, the ‘x’ in the sigmoid function is denoted by the weighted sum of the variables, the illustration below further explains how this process works</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/602/0*gUsaFvek3AhvQe1d.png" /></figure><h3>K-Nearest Neighbors</h3><p>In KNN the value of a data point is determined by the value around it, it works on a majority wins principle, If there 5 nearest neighbors and 3 of them belong to Class A, one belongs to Class B and one belongs to Class C, the unknown data point will be classified as belonging to Class A. The K in the K nearest neighbors algorithm gives the number of neighbors we must consider, so if K=5 you will consider 5 neighbors, Now that we have understood the K of KNN let’s understand how the nearest neighbors are calculated</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/528/0*Ef4KlcNKpcW7uVfh.gif" /></figure><p>There are several ways of identifying the nearest neighbors I’ll be explaining the 2 most common ones</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/320/0*IckzchqnL4dJQK84.jpg" /></figure><p><strong>Manhattan Distance</strong></p><p>Manhattan Distance is calculated using the absolute sum of differences of Cartesian co-ordinates</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/509/0*fzsPnAPXko5BS40J.png" /></figure><p><strong>Euclidean Distance</strong></p><p>Euclidean Distance is calculated using the square of the sum of differences of Cartesian co-ordinates, in the Sklearn libraries this is the default metric for calculating KNN.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*RsA6_3JMvDtcvI_a.png" /></figure><p>Now there’s no way to know what’s the best value of K is in a dataset, so the best way to determine what gives the best accuracy is put the values in a loop and plot the K-values vs. Accuracy Curve</p><h3>Naïve Bayes</h3><p>There are many types of Naïve Bayes Algorithms, I’ll be focusing on Multinomial Naïve Bayes Classification, the other types of Naïve Bayes will be covered in a separate article</p><p>The Naïve Bayes Algorithm is modelled around the Bayes theorem, the below figure calculates the probability of A happening provided that B has occurred, the Naïve Bayes Classifier assumes that all attributes contribute equally to the outcome, i.e. they all have the same weight no attribute is more important or less important, this is why we call it “Naïve”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/468/0*r2zyQWHCwe8EiQIa.jpeg" /></figure><h3>Decision Tree Classification</h3><p>Decision Tree as the name suggests is a tree that makes decisions, this algorithm follows a rule based approach. What the algorithm does is, it uses all the features to create yes/no questions and continues to split the decisions, until all the datapoints are isolated in their respective classes, the first node in this tree is called the root and the bottom-most node is called the leaf node</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/542/0*L97Qj0L3mi1GVbH7.png" /></figure><p>Now how do you create these questions, essentially how do you split a tree, to do this we use a loss function that compares distribution pre and post splitting, i.e. Gini Impurity &amp; Entropy, the goal is to minimize these functions in order to get a better split.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jNS3Tvu-DvURiySc" /></figure><p><em>Next time I’ll be writing about Recommendation Systems, if you like this article please leave an upvote, if you don’t like it please let me know what I should work on :)</em></p><p><a href="https://medium.com/mlearning-ai/mlearning-ai-submission-suggestions-b51e2b130bfb">Mlearning.ai Submission Suggestions</a></p><p>🔵 <a href="https://medium.com/mlearning-ai/mlearning-ai-submission-suggestions-b51e2b130bfb"><strong>Become a Writer</strong></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ee6dff9c2af5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Machine Learning 101: All things Regression]]></title>
            <link>https://nicolemichelledsouza.medium.com/machine-learning-101-all-things-regression-43372b662ed4?source=rss-7fe846709991------2</link>
            <guid isPermaLink="false">https://medium.com/p/43372b662ed4</guid>
            <category><![CDATA[lasso-regression]]></category>
            <category><![CDATA[ridge-regression]]></category>
            <category><![CDATA[regression]]></category>
            <category><![CDATA[linear-regression]]></category>
            <category><![CDATA[elastic-net]]></category>
            <dc:creator><![CDATA[Nicole Michelle]]></dc:creator>
            <pubDate>Tue, 01 Feb 2022 12:38:58 GMT</pubDate>
            <atom:updated>2022-02-01T12:38:58.251Z</atom:updated>
            <content:encoded><![CDATA[<p>In the last article I introduced machine learning as a concept and spoke about the different types of learning, this time we will look at Regression and it’s various types</p><p>Regression is a supervised learning algorithm, it is used to predict continuous values like housing prices, weather forecasts, but fundamentally it identifies the relations between variables and uses this understanding to provide an output. Anyone that takes a course in Machine Learning is first introduced to linear regression because of it’s simplicity and understandability , along with linear regression we will also cover some other types of regression, there are several other algorithms excluding these for which I’ll do a part 2 soon</p><p>Let’s begin!</p><h3><strong><em>Linear Regression</em></strong></h3><p>Linear Regression as the name suggest assumes that the relation between input x and output y is linear, in simple linear regression a simple linear input variable is mapped to a single linear output variable, when there are multiple input variables it is called multiple linear regression. The linear regression is a very simple slope equation</p><p><strong><em>y = a₀ +a₁*x</em></strong></p><p>where y is the dependent variable, the one we predict, for example how much salary will you make after x year’s of experience. x is the independent variable that causes y to change, in this case i.e. in simple linear regression there is only one independent variable, in multiple linear regression we find more than one independent variable. <strong>a₁</strong> is simply the coefficient it can be taught of as the multiplier. The figure before illustrates how the model tries to fit the data better with each iteration</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/864/0*CT8mdO9FIHFokeF8.gif" /><figcaption><a href="https://dphi.tech/blog/tutorial-on-linear-regression-using-gradient-descent/">Image Credit</a></figcaption></figure><h3><strong><em>Polynomial Regression</em></strong></h3><p>Now linear regression as the name suggests depends on the relation between x and y being linear, however if the data is non linear, the algorithm won’t work! This is where polynomial regression comes in, it fits the models better by assuming the non linear relation between x and y, instead of a linear equation being formed we get a quadratic equation or a cubic equation depending on the degree m.</p><h3>y = a₀ +a₁*x+ a₁*x² + a₃*x³+ …..aₘ*xᵐ</h3><p>The figure below illustrates how linear regression fails to fit to the data because of it’s non linear nature, while polynomial regression does a fantastic job!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/0*5zC7_MNGQcaEA4jH.png" /></figure><h3><strong><em>Ridge Regression (L2)</em></strong></h3><p>In regression sometimes the model fit too closely to the dataset i.e. they overfit. Ridge regression works on solving the problem of overfitting by trading a model that fits the data perfectly with a model that provides better predictions in the long term, it achieves this by putting a penalty on the weights on the traditional least squares method, however it’s important to understand that without the small bias that ridge regression creates the regression model will have a very high variance</p><p>The figure below has in green the formula used to calculate regression known as the Residuals Sum of Squares, the red text is the loss function that’s added in ridge regression, it penalizes the model for wrong predictions which is how the overall prediction of the model is improved</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/857/0*zafPvxxCXZS7-yx9.png" /></figure><h3><em>Lasso Regression(L1)</em></h3><p>Lasso Regression is very very similar to Ridge Regression in a way that both of them work towards avoiding overfitting, however the way they do it is a little different, in ridge regression we square the slope(<strong>β</strong>ⱼ) in the loss function, but in lasso regression we take it’s absolute value.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/364/0*BNGLEMcUfINyX3ye.png" /></figure><p>Like Ridge Regression, Lasso Regression too has a very small bias but much lesser variance than the least squares method. So what’s the difference?</p><p>Ridge Regression can shrink the slope only asymptotically to 0,whereas as Lasso Regression can shrink slope all the way to 0.Simply put let’s assume you need to calculate the price of your used car and one of the features is your astrological sign (something that has no relation to used cars),now ridge regression can greatly shrink the value of this parameter to give you a better prediction, it works best when most of it’s variables are useful as it cannot eliminate them, but Lasso Regression can completely eliminate that parameter which consequently gives you an even better prediction!</p><h3>ElasticNet Regression</h3><p>Ridge Regression is great when most of the variables provided are useful, else we should go ahead with Lasso Regression. However when provided with a dataset of hundreds of parameters it becomes difficult to know which parameters are important and which are not!</p><p>This is where ElasticNet comes into the picture, it essentially works towards giving you the best of both worlds</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/731/0*JHw3HYtkIUsTY0mM.png" /></figure><p>i.e. Ridge Regression &amp; Lasso Regression, by combining the loss functions of both and using cross validation to determine the best L1 and L2 values. The ElasticNet equation works especially well when there are correlations between parameters</p><p><em>Next time I’ll be writing about Classification Algorithms, if you like this article please leave an upvote, if you don’t like it please let me know what I should work on :)</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=43372b662ed4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Downloading your LinkedIn Connections Data]]></title>
            <link>https://nicolemichelledsouza.medium.com/downloading-your-linkedin-connections-data-f76c6e61a9aa?source=rss-7fe846709991------2</link>
            <guid isPermaLink="false">https://medium.com/p/f76c6e61a9aa</guid>
            <category><![CDATA[dashboard]]></category>
            <category><![CDATA[linkedin]]></category>
            <category><![CDATA[dataset]]></category>
            <category><![CDATA[tableau]]></category>
            <dc:creator><![CDATA[Nicole Michelle]]></dc:creator>
            <pubDate>Sun, 30 Jan 2022 16:35:29 GMT</pubDate>
            <atom:updated>2022-01-30T16:35:29.897Z</atom:updated>
            <content:encoded><![CDATA[<p>In this article I will show you how to download your LinkedIn Connections Data. So let’s get started!</p><ol><li>Click on the my Network Section at the top of your LinkedIn Page</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZhfQAe-Zj2veiMk-eHLOZw.png" /></figure><p>2. On the left side of the page, you will find an option labelled Connections, click on that</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/497/1*DtNfOWQfhvUygLdOhnwkDQ.png" /></figure><p>3. Click on the Manage Synced and Import Contacts on the top right</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/695/1*zkAA_WVjdAskoHYqie4LVw.png" /></figure><p>4. Under the Advanced Options, select Export contacts</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8loyfisRL3H_QaAVM7OVww.png" /></figure><p>5. Under the list on the left side of the page select Data Privacy and then click on How LinkedIn uses you data</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*km0-Lnqx0nmTHNvIldpP9w.png" /></figure><p>6. Click on get a copy of your data, you can either choose to download your entire data archive or select what specific csv files you want</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*33dRmSi48ixFgn0APJdi1g.png" /></figure><p>7. For the purpose of this tutorial you will only need Connections &amp; Invitations, select the options needed and click on <em>Request Archive, </em>after which you will be asked to enter your password and then in about 10 mins you will receive an email containing you data archive</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mrteTOr4_xTRkh0D6qkesw.png" /></figure><p>8. Extract the csv files from the folder, in the Connections.csv the first 3 rows are a disclaimer we shall delete that</p><p>9. We will also get rid of the email column as it is empty (because of GDPR norms)</p><p>That’s all on how to get the dataset!</p><p>You can use this for a variety of things like sentiment analysis of your messages ,or maybe as a data analytics project</p><p>This how I have used the dataset : <a href="https://public.tableau.com/views/LinkedInConnectionsAnalysis/Dashboard1?:language=en-US&amp;:display_count=n&amp;:origin=viz_share_link">https://public.tableau.com/views/LinkedInConnectionsAnalysis/Dashboard1?:language=en-US&amp;:display_count=n&amp;:origin=viz_share_link</a></p><p><em>If you enjoyed reading this article please like it, if you didn’t let me know how I can improve :)</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f76c6e61a9aa" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Machine Learning 101:Introduction]]></title>
            <link>https://nicolemichelledsouza.medium.com/machine-learning-101-97960e68f4c7?source=rss-7fe846709991------2</link>
            <guid isPermaLink="false">https://medium.com/p/97960e68f4c7</guid>
            <category><![CDATA[regression]]></category>
            <category><![CDATA[reinforcement-learning]]></category>
            <category><![CDATA[machine-intelligence]]></category>
            <category><![CDATA[classification]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Nicole Michelle]]></dc:creator>
            <pubDate>Thu, 23 Dec 2021 17:25:59 GMT</pubDate>
            <atom:updated>2021-12-31T04:15:30.434Z</atom:updated>
            <content:encoded><![CDATA[<p>While Machine Learning has been around since the 50s starting with Arthur Samuel’s famous checkers game, it has recently gained traction because of the large number of libraries and the ease of using those libraries, more and more businesses have been using data to generate valuable insights and everyone seems to be talking about Machine Learning, but what is machine learning? Machine Learning can be defined as</p><p><strong><em>“A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.”- Tom Mitchell</em></strong></p><p>Machine Learning is broadly classified into 3 types, namely</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jZ_cfq1sp2vJudhEHC-dJw.jpeg" /><figcaption><a href="https://luminousmen.com/post/machine-learning-types/">luminousmen</a></figcaption></figure><h3>Supervised Learning</h3><p>In Supervised Learning, the agent already knows what the right answer looks like and it establishes the relation between the input data and output which helps the agent make predictions. Supervised Learning has 2 types: Regression where the data predicted here has a continuous output and Classification where the data predicted has a discrete output, if you had to predict whether or not you can buy a house that would be classification, if you had to predict how much a house would cost you that would be regression.</p><h3>Unsupervised Learning</h3><p>In Unsupervised Learning, the agent doesn’t know what the right answer looks like, it identifies the relationship between data points and helps give the user several hidden insights from the data. Unsupervised Learning is further classified into Clustering where data is grouped together into clusters based on similarities and Association where the agent determines which items in the dataset are more likely to occur together if the agent groups customers together based on consumption/spending habits that would be clustering and if the agent identifies which items are more likely to be bought together eg. peanut butter &amp; jelly that’s association</p><h3>Reinforcement Learning</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/543/1*lYwzSb52VAVgxnopTVcmIg.jpeg" /><figcaption><a href="https://www.kdnuggets.com/2019/10/mathworks-reinforcement-learning.html">KD nuggets</a></figcaption></figure><p>Reinforcement Learning is nothing like the other 2 types, supervised learning made use of labeled data to make predictions, unsupervised learning helped capture the relations between data, reinforcement learning makes use of rewards and punishments, the goal of this method is to maximize the rewards, think of it like training a dog when it does something right you say good dog and toss him a treat when it does something wrong you say bad dog!</p><p><em>If you liked this article please leave an upvote, I will be writing about regression next week</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=97960e68f4c7" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>