<?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 Divyashie Soopal on Medium]]></title>
        <description><![CDATA[Stories by Divyashie Soopal on Medium]]></description>
        <link>https://medium.com/@divyashie16?source=rss-6a774a35c51c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*rVjFAKVhij7RUYkyFlQxeA.jpeg</url>
            <title>Stories by Divyashie Soopal on Medium</title>
            <link>https://medium.com/@divyashie16?source=rss-6a774a35c51c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 09:03:59 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@divyashie16/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[101: Google Sheets API in Streamlit]]></title>
            <link>https://divyashie16.medium.com/101-google-sheets-api-in-streamlit-ab4e9468d551?source=rss-6a774a35c51c------2</link>
            <guid isPermaLink="false">https://medium.com/p/ab4e9468d551</guid>
            <category><![CDATA[streamlit]]></category>
            <category><![CDATA[python]]></category>
            <category><![CDATA[google-sheets-api]]></category>
            <dc:creator><![CDATA[Divyashie Soopal]]></dc:creator>
            <pubDate>Mon, 23 Jan 2023 20:02:28 GMT</pubDate>
            <atom:updated>2023-01-23T20:02:28.292Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UsXgdTvo94P5oZqwXEMC9Q.jpeg" /><figcaption>Google Sheets + Streamlit [.jpeg]</figcaption></figure><p><a href="http://streamlit.io/">Streamlit</a> is by far the fastest way I have built and shared data apps. It is written in pure Python language. To top it up, you can also embed front-end technologies such as HTML.</p><p><em>For my use case, I have built a simple web app that allows users to enter predicted scores for the Fifa world cup of 2022. The predicted scores were stored in a google sheet. I would then manually update the actual score for the match. And, then if the user correctly predicted the exact score for winning the match they will receive a total of 3 points. If the user predicted the right winning team, even though they fail to predict the actual winning scores, they receive a total of 1 point. Else, the default is zero points. Pretty much imitates the conventional scoring of Fifa prediction matches.</em></p><p>There is a nice tutorial that I followed to get started with <a href="https://aryanirani123.medium.com/read-and-write-data-in-google-sheets-using-python-and-the-google-sheets-api-6e206a242f20">how to read and write in google sheets using python</a>. Below is the skeleton guideline to help you get started.</p><ul><li>Create a Google Cloud Console project with Google Sheets API enabled</li><li>Create a Service Account (credentials) that will be used to interact with the Google Sheet</li><li>Get credentials (download JSON key file) and then add it to <a href="https://blog.streamlit.io/secrets-in-sharing-apps/">a secrets management file inside your project directory</a>.</li><li>Install GSpread library with pip</li><li>Add service account email to your Google Sheet</li><li>Start writing your Streamlit python script</li></ul><p>once you have created a new project on the Google Cloud Console and you have enabled the Google Sheets API to it, you can then proceed to implement it inside your main project file from which Streamlit executes.</p><h4>An example of how I implemented the Google Sheet API in Streamlit</h4><p>Besides, installing the required library, I also used `<strong>gspread_pandas</strong>` to save the data inside of the data frame for ease of manipulation. The scope that I used contains the list of URLs for spreadsheets and google drive APIs. Now, using a service account and passing the API secret token, I created a unique credential. Then, I used this credential to create a client session.</p><p>With the client session and the Google spreadsheet, I was able to read all the current records in a data frame.</p><pre>from google.oauth2 import service_account<br>from gsheetsdb import connect<br>from gspread_pandas import Spread,Client</pre><pre>scope=[<br>         &quot;https://www.googleapis.com/auth/spreadsheets&quot;, &#39;https://www.googleapis.com/auth/drive&#39;<br>]<br>credentials = service_account.Credentials.from_service_account_info(<br>                st.secrets[&quot;gcp_service_account&quot;], scopes = scope)<br>client = Client(scope=scope,creds=credentials)<br>spreadsheetname = &quot;Database&quot;<br>spread = Spread(spreadsheetname,client = client)<br>#st.write(spread.url)<br><br>sh = client.open(spreadsheetname) </pre><pre>def loadSpreadsheet(sheet): <br>    worksheet= sh.worksheet(sheet)<br>    df = pd.DataFrame(worksheet.get_all_records())<br>    return df <br><br>def updateSpreadsheet(sheet, dataframe, col): <br>    spread.df_to_sheet(dataframe[col], sheet=sheet, index=False)</pre><p>After reading the spreadsheet, I can manipulate this data frame for further assigning the actual scores of the matches. Once, I had manually inserted the actual winning scores of the Fifa match, this new information would simply be updated using the <strong>updateSpreadsheet </strong>function. This technique is all about reading and writing data in real-time to a Google Sheet.</p><p>Here is a sneak peek at how I implemented the load and update functions:</p><pre>def arrayCountries(): <br>    matches = { &#39;Croatia&#39; : &#39;Morocco&#39;, &#39;Argentina&#39; : &#39;France&#39;} #change<br>    res = []<br>    col = [&#39;name&#39;, &#39;match1&#39;, &#39;score1&#39;, &#39;match2&#39;, &#39;score2&#39;]<br>    with st.form(&#39;Test&#39;): <br>        name = st.text_area(&quot;Entrez le nom de l&#39;équipe&quot;, key=&#39;text_key&#39;)<br>        for match1, match2 in matches.items(): <br>            score1, score2 = match(match1,match2)<br>            result = {&#39;name&#39;: name, &#39;match1&#39;:match1, &#39;score1&#39;: score1, &#39;match2&#39;: match2, &#39;score2&#39;: score2}<br>            res.append(result)<br>        submitButton = st.form_submit_button(&quot;Envoyer ⚽&quot;)<br>        if submitButton: <br>            sheet = &quot;sheet1&quot; <br>            current_df = pd.DataFrame(res)<br>            print(current_df)<br>            df = loadSpreadsheet(sheet)<br>            new_df = df.append(current_df, ignore_index=True)<br>            updateSpreadsheet(sheet, new_df, col)<br>            st.markdown(&quot;&lt;h4&gt;Merci pour votre réponse&lt;/h4&gt;&quot;,unsafe_allow_html=True)</pre><p>I created a dictionary to hold the name of the matches that changes based on the frequency of the Fifa matches. Using a Streamlit in-built form functionality, I was able to detect any change coming through and then pass that to the <strong>updateSpreadsheet</strong> function.</p><p><em>If you are curious about the entire code base, feel free to check out my </em><a href="https://github.com/divyashie/fifa-2022-final"><em>Github repository</em></a><em> on it.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ab4e9468d551" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Secret Santa by Email]]></title>
            <link>https://divyashie16.medium.com/secret-santa-by-email-1f35f4b8f32?source=rss-6a774a35c51c------2</link>
            <guid isPermaLink="false">https://medium.com/p/1f35f4b8f32</guid>
            <category><![CDATA[outlook]]></category>
            <category><![CDATA[christmas]]></category>
            <category><![CDATA[python]]></category>
            <category><![CDATA[email-automation]]></category>
            <dc:creator><![CDATA[Divyashie Soopal]]></dc:creator>
            <pubDate>Tue, 03 Jan 2023 14:19:30 GMT</pubDate>
            <atom:updated>2023-01-03T14:19:30.592Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/740/1*o4S2orraYunY_QQMV9-6KQ.jpeg" /><figcaption>Ref: [Fun blog]</figcaption></figure><p>During the holiday season, my team decided we will gift each other a present. However, we wanted it to be more fun. A secret. Closer to the concept of Secret Santa.</p><p>Traditionally, to execute our Secret Santa project, we would write names on a piece of paper, fold them, keep it in a big box, shuffle the papers, and then generate a name/the chosen person here. But, there is a catch. What if the same person gets his/her own name? We could not afford that.</p><p>We wanted a more efficient, time-saving, error-free system, but also quick to develop. What if each person gets emailed their respective wish for whom they will act as Santa? This whole process will be automated using a simple logic of name shuffling.</p><p>Also, to address the issue of name repetition mentioned above, I simply created a duplicate store for the names and remove those that were already assigned their (wish). This code assumes that we have an even list of people. If we wanted to do a one-to-many relationship, that is, one Santa many kids we would have to slightly re-write this logic.</p><pre>def santaKid(team_name): <br>    &quot;&quot;&quot;<br>    generates a list of dictionary containing unique kid to a santa. <br>    &quot;&quot;&quot;<br>    copy_team = team_name.copy()<br>    result = []<br>    for santa in team_name: <br>        kid = random.choice(copy_team)<br>        while santa == kid: <br>            kid = random.choice(copy_team)<br>        result.append({santa: kid})<br>        copy_team.remove(kid)<br>    return result </pre><p>The next step was to create a stylish Christmas-themed email. For that, I used HTML to style the content of the email as follows. <em>The message is written in French — the language used by the Company I work for</em>.</p><pre>def emailContent(santa, kid): <br>    message = f&#39;&#39;&#39;&#39;<br>        &lt;!DOCTYPE html&gt;<br>        &lt;html&gt;<br>            &lt;body&gt;<br>                &lt;div style=&quot;background-color:#ff7f50;padding:10px 20px;&quot;&gt;<br>                    &lt;h2 style=&quot;font-family:&#39;Mountains of Christmas&#39;, cursive;color#ffa07a; text-align:center;&quot;&gt; Echange des cadeaux chez Apsim&lt;/h2&gt;<br>                &lt;/div&gt;<br>                &lt;div style=&quot;padding:20px 0px&quot;&gt;<br>                    &lt;div&gt;<br>                        &lt;img src=&quot;https://www.dukeshillham.co.uk/Portals/0/product/images/secret_satnta_2019.jpg&quot; style=&quot;width:100%;&quot;&gt;<br>                        &lt;div style=&quot;text-align:center;&quot;&gt;<br>                            &lt;h3&gt;{santa}&lt;/h3&gt;<br>                            &lt;p&gt;Ho Ho Ho!! &lt;/p&gt; <br>                            &lt;p&gt; Histoire de faire perdurer l&#39;esprit de noël dans nos bureaux, tu es invité à participer à l&#39;échange des cadeaux.&lt;/p&gt;<br>                            &lt;p&gt; L&#39;elfe de noël t&#39;a choisi pour être le secret Santa de &lt;b&gt;{kid}&lt;/b&gt;! {kid} a été bon enfant et attend son cadeau avec patience.&lt;/p&gt;<br>                            &lt;img src= &quot;https://clipground.com/images/christmas-box-clipart-13.jpg&quot; style=&quot;height:300px;&quot;&gt;<br>                        &lt;/div&gt;<br>                    &lt;/div&gt;<br>                &lt;/div&gt;<br>                &lt;link rel=&quot;preconnect&quot; href=&quot;https://fonts.googleapis.com&quot;&gt;<br>                &lt;link rel=&quot;preconnect&quot; href=&quot;https://fonts.gstatic.com&quot; crossorigin&gt;<br>                &lt;link href=&quot;https://fonts.googleapis.com/css2?family=Mountains+of+Christmas:wght@700&amp;display=swap&quot; rel=&quot;stylesheet&quot;&gt;<br>            &lt;/body&gt;<br>        &lt;/html&gt;<br>        &#39;&#39;&#39;<br>    return message </pre><p>The Email’s display can be viewed here: <a href="http://tpcg.io/_IAX4H9">http://tpcg.io/_IAX4H9</a>.</p><p>Finally, to automate sending out the Email, I used the smtplib. This is a neat example of it:</p><pre>def automateEmail(recipient,message): <br>    msg = EmailMessage()<br>    msg[&#39;Subject&#39;] = &#39;Mission: Santa Secret 🎅&#39;<br>    msg[&#39;From&#39;] = sender <br>    msg[&#39;To&#39;] = recipient<br>    msg.set_content(message, subtype=&#39;html&#39;)<br><br>    try: <br>        smtp = smtplib.SMTP(&quot;smtp-mail.outlook.com&quot;, port=587) <br>        smtp.starttls()<br>        smtp.login(sender, password)<br>        smtp.send_message(msg)<br>        smtp.quit()<br>        print(&quot;Email sent successfully!&quot;)<br>    except Exception as e: <br>        print(e)</pre><p>Please note that for Outlook the port is 587 but it will be different for other e-mail providers.</p><p><em>The next time you need to conduct a draw company-wide, you will try to dev it ;)</em></p><p><strong>Reference:</strong></p><p><a href="https://github.com/Apsim-Ltd/secret_santa.git">https://github.com/Apsim-Ltd/secret_santa.git</a></p><p><a href="https://www.fun.com/blog/p-1252-whos-your-secret-santa.aspx">https://www.fun.com/blog/p-1252-whos-your-secret-santa.aspx</a></p><p><a href="https://docs.python.org/3/library/smtplib.html">smtplib - SMTP protocol client</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1f35f4b8f32" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Airline Data Analysis Using Spark Technologies]]></title>
            <link>https://divyashie16.medium.com/airline-data-analysis-using-spark-technologies-d87f756328c2?source=rss-6a774a35c51c------2</link>
            <guid isPermaLink="false">https://medium.com/p/d87f756328c2</guid>
            <dc:creator><![CDATA[Divyashie Soopal]]></dc:creator>
            <pubDate>Sat, 13 Jun 2020 03:22:39 GMT</pubDate>
            <atom:updated>2020-06-13T03:22:39.445Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/300/1*r_bPesqkhx56AOXGbTBCxw.jpeg" /></figure><p>This post is about how I navigated my way through PySpark in general.</p><p>I analyzed the US domestic flight dataset using PySpark dataframes. The aim of the project was to predict which flight/flight carrier is most likely to be cancelled or delayed.</p><h3>Introduction</h3><p>The dataset was taken from Kaggle, comprised 7 CSV files containing data from 2009 to 2015, and was about 7GB in size. It contained information about airlines, delay information, location details (origins and destinations), and cancellation (reasons labelled as cancellation codes). It also had certain technical details such as the time that the plane was stationary on ground.</p><p>My approach revolved around Apache Spark , specifically using PySpark with AWS EMR cluster support. This virtual set-up uses distributed, in-memory data structures to improve speeds for many data processing workloads by several orders of magnitude. As a unified analytics engine, it allows us to parallelize large data processing tasks on a distributed cluster.</p><p>PySpark is also great for exploratory data analysis at scale, building machine learning pipelines, and creating ETLs (extract, transform, load) tools. Once the EMR cluster was “running,” I connected to Jupyterlab available as notebooks under AWS services and loaded the data from S3 buckets by importing it through the Kaggle website. I performed exploratory data analysis and model prediction.</p><h3>Data Profiling</h3><p>To get a quick glance at the data, we loaded the 2009 csv file into a Resilient Distributed Dataset (RDD) which is Spark’s representation of a dataset that is distributed across RAM, in the memory of a cluster of many machines. Through this Spark session, I found that there was about 27 unnamed variables and multiple null values.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*oC5aTuAVWp2h4coO" /><figcaption>Quick glance at the 2009 csv in pyspark dataframe</figcaption></figure><p>After combining the dataset (2009 to 2015 data), I created 3 major categorical CSV files namely: flight-cancellation, flight-delays, and flight-profile. This was essential to debug.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*T8A-s-EKiNvY-Aqt" /><figcaption>Setup for s3 bucket</figcaption></figure><h3>Data Wrangling</h3><ol><li><strong>Data Cleaning and Preparation</strong></li></ol><p>Originally, there was 28 variables to work with. After removing the unnamed columns, the remaining 19 were checked for null values (as shown below). In addition, only columns with relevant data about the flight profile, delay, and cancellation information were kept.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/508/0*MwvEl2YYzVkgeGBz" /><figcaption>Exploring null values throughout dataset</figcaption></figure><p>With time sensitive data, it was hard to recreate the null values with either mean or median of the data, so I had to drop these records which boils down to over 6+ million records to still work with.</p><p>After cleaning out the date column, it contained only the yearly information which was used in the analysis.</p><p><strong>2. Statistical Data Analysis</strong></p><ul><li>From the profile sub-dataset, there were 61,556,964 flight profile records including 7605 US domestic flights in total, around 380 unique origins and 378 unique destinations. This information was populated in temporary tables inside the Spark session to query specifics.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/462/0*Y0Ach7t4mACeeZ9w" /><figcaption>Glance at profile sub-dataset</figcaption></figure><ul><li>From the delay sub-dataset, there were negative values in the departure delay column indicating that around 3,1204,918 flights were actually ahead of their scheduled time; that is roughly 50% of the dataset. This means that 50% of the time flights were delayed.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/482/0*C9IAYgs1dOowOwvv" /><figcaption>Glance at delay sub-dataset</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/527/0*AHfOm493Yw6-TMcU" /><figcaption>Ahead of time (&lt;0), On time (0), Delay (&gt;0)</figcaption></figure><ul><li>Within the cancellation sub-dataset, the cancel column is categorical data represented by either a 0 or a 1 (not cancelled or cancelled respectively).The reasons for cancellation from the data are as follows:</li></ul><p><em>A — Airline/Carrier (315,450)</em></p><p><em>B — Weather (476,446)</em></p><p><em>C — National Air System (180,605)</em></p><p><em>D — Security (708)</em></p><p>The most reported reason was weather and the least reason was security.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/483/0*Ah2PQuATJcOJcQwo" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/478/0*dPkEbiDjG6spqheS" /></figure><p>YV (Mesa Airlines), F9 (Frontier Airlines), and HA (Hawaiian Airlines) were among the 10 top airlines with the most flight operations from 2009 to 2015 (Figure 1).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/696/0*bQ6sXZKU0Cxb7i07" /><figcaption>Fig 1: Top 10 Airlines with the most flight operations from 2009 to 2015</figcaption></figure><p>In the analysis, 2013 had the most flights, 2015 showed the lowest one. Overall, there were slight variations with the total number of flights per year; that is the number stayed in between 6 to 6.5 millions (Figure 2).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*e15VpnyLHM-LECdK" /><figcaption>Fig 2: Number of flights across the years 2009 to 2015</figcaption></figure><p>The proportion for the total flight cancellation reasons across 2009 to 2015 is shown below (Figure 3).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/717/0*iydmBvaqtWm3sKR4" /><figcaption>Fig 3: Proportion of flight cancellation reasons across the years 2009 to 2015</figcaption></figure><h3><strong>Model Prediction</strong></h3><p>To predict whether the flight would be cancelled or not, the dataset was modeled as a binary classification problem with categorical variables.</p><ol><li>Prepare the data for machine learning: use StringIndexer, OneHotEncoder, and VectorAssembler to transform our features</li><li>Split the data into a 70/30 test and train ratio</li><li>Apply models: logistic regression, decision tree classifier, random forest and gradient boosted trees</li><li>Compare all the models’ accuracy for predicting cancellations</li></ol><p><strong>For the logistic regression model:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/425/0*X2fbkUi2fEvPmCzA" /><figcaption>Beta coefficient estimates for our parameters</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/442/0*ExmU_J_XFQTiv6eA" /><figcaption>True positive vs false positive rate for the ROC curve</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/410/0*cka_A36RIPzui3Qj" /><figcaption>Precision and Recall for logistic regression</figcaption></figure><p>Prediction accuracy was evaluated using the BinaryClassificationEvaluator.</p><p><strong>For the RandomForest model:</strong></p><p>Decision Tree Classifiers can handle categorical variables well and are able to capture non-linearity. With the aid of the pyspark.ml.classification library, the DecisionTreeClassifier was imported. This was essential to show the probability of each prediction and evaluate the model.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/611/0*_es6aaHaMFI0fEUS" /></figure><p><strong>For the gradient-boosted classifiers:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/619/0*Q65tiTr8rgMDfHdF" /></figure><h3>Results</h3><p>From the initial tests, the logistic regression model and gradient boosted classifiers outperformed random forest.</p><p>To test out various hyperparameters, for the gradient boosted classifier and to attempt an improvement in the performance, ParamGridBuilder and CrossValidator were used to find the maxDepth and maxBin that gave out the best accuracy.</p><p>With cross validation, the best parameters for a gradient boosted tree are a maximum depth of 6 and maximum bins of 20. With these improvements, the accuracy for Gradient Boosted Tree (GBT) changed to 63.8%.</p><p>From the model, the airline with the highest probability of being cancelled was ExpressJet Airlines (EV), with Delta Airlines (DL) having the least probability of being cancelled.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/553/0*fyS8r5xw2NlSuWwN" /><figcaption>Lowest Probability of Cancellation: Delta Airlines</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/555/0*zE-o--yyN6pmv9Fs" /><figcaption>Highest Probability of Cancellation: ExpressJet Airlines</figcaption></figure><p>A schema for the flight delays dataset was created with feature and label column. Similarly, with the cancellation classification, VectorAssembler and StringIndexer were used to transform the dataset. The linear regression model was applied to get the coefficients and intercepts; they were used to test on the linear regression and decision tree models.</p><p>For both models, high RMSE values were obtained, implying that the airline is not a meaningful predictor of delay time.</p><h3>Conclusions</h3><p>The project was able to identify whether the airline is likely to cancel with an accuracy of 63%. GBT performed the best for this case. From the data, the most likely flight to be cancelled was ExpressJet Airlines and the least likely was Delta Airlines.</p><p>However, airline identity was not a meaningful predictor of delay time. This suggests that delay of flight is likely to be for reasons other than the airline, such as weather conditions or airport.</p><h3><strong>References</strong></h3><p><strong>[1] </strong>Bureau of Transportation Statistics. (2016). Airline On-Time Performance and Causes of Flight Delays. Retrieved from <a href="https://catalog.data.gov/dataset/airline-on-time-performance-and-causes-of-flight-delays-on-time-data">https://catalog.data.gov/dataset/airline-on-time-performance-and-causes-of-flight-delays-on-time-data</a></p><p><strong>[2] </strong>Deshpande, V., &amp; Arikan, M. (2011). The Impact of Airline Flight Schedules on Flight Delays. <em>Manufacturing &amp; Service Operations Management, 14</em>, 423–440. Retrieved from <a href="https://pubsonline.informs.org/doi/10.1287/msom.1120.0379">https://pubsonline.informs.org/doi/10.1287/msom.1120.0379</a></p><p><strong>[3] </strong>Mu, Y. (2019, August). Airline Delay and Cancellation Data, 2009–2018. Retrieved April 2020 from <a href="https://www.kaggle.com/yuanyuwendymu/airline-delay-and-cancellation-data-2009-2018/data">https://www.kaggle.com/yuanyuwendymu/airline-delay-and-cancellation-data-2009-2018/data</a></p><p><strong>[4] </strong>Chakrabarty, Navoneel, et al. ”Flight Arrival Delay Prediction Using Gradient Boosting Classifier.” Emerging Technologies in Data Mining and Information Security. Springer, Singapore, 2019. 651–659. Retrieved from <a href="https://www.researchgate.net/publication/327389509_Flight_Arrival_Delay_Prediction_Using_Gradient_Boosting_Classifier">https://www.researchgate.net/publication/327389509_Flight_Arrival_Delay_Prediction_Using_Gradient_Boosting_Classifier</a></p><p><strong>[5]</strong> Yi Ding ”Predicting flight delay based on multiple linear regression”, IOP Conference Series: Earth and Environmental Science. Retrieved from <a href="https://iopscience.iop.org/article/10.1088/1755-1315/81/1/012198">https://iopscience.iop.org/article/10.1088/1755-1315/81/1/012198</a></p><p><strong>[6] </strong>Belcastro, L. &amp; Marozzo, Fabrizio &amp; Talia, Domenico &amp; Trunfio, Paolo. (2016). Using Scalable Data Mining for Predicting Flight Delays. ACM Transactions on Intelligent Systems and Technology. 8. 10.1145/2888402. Retrieved from <a href="https://dl.acm.org/doi/10.1145/2888402">https://dl.acm.org/doi/10.1145/2888402</a></p><p><strong>[7] </strong>Li, S. “Machine Learning with PySpark and MLlib — Solving a Binary Classification Problem”. Towards Data Science. Retrieved from</p><p><a href="https://towardsdatascience.com/machine-learning-with-pyspark-and-mllib-solving-a-binary-classification-problem-96396065d2aa">https://towardsdatascience.com/machine-learning-with-pyspark-and-mllib-solving-a-binary-classification-problem-96396065d2aa</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d87f756328c2" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>