<?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 Temiloluwa Akintunde on Medium]]></title>
        <description><![CDATA[Stories by Temiloluwa Akintunde on Medium]]></description>
        <link>https://medium.com/@akintundetemiloluwa.a?source=rss-cc2a9b4c51bb------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*inad2aq1HeLSoRLFnCubfQ.jpeg</url>
            <title>Stories by Temiloluwa Akintunde on Medium</title>
            <link>https://medium.com/@akintundetemiloluwa.a?source=rss-cc2a9b4c51bb------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 14:48:21 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@akintundetemiloluwa.a/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[CREDIT CARD FRAUD PREDICTION ANALYSIS]]></title>
            <link>https://medium.com/@akintundetemiloluwa.a/credit-card-fraud-prediction-analysis-3ba9cd865310?source=rss-cc2a9b4c51bb------2</link>
            <guid isPermaLink="false">https://medium.com/p/3ba9cd865310</guid>
            <category><![CDATA[credit-card-fraud]]></category>
            <category><![CDATA[anomaly-detection]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Temiloluwa Akintunde]]></dc:creator>
            <pubDate>Wed, 15 Jan 2025 12:42:13 GMT</pubDate>
            <atom:updated>2025-01-15T12:42:13.769Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*umjH2yd9OONmJLRPyVfBtQ.jpeg" /></figure><h3>Introduction</h3><p>Credit card fraud is a significant challenge in the financial sector, leading to substantial financial losses and undermining customer trust. With the increasing volume of digital transactions, it has become critical for organizations to adopt advanced techniques to detect and prevent fraudulent activities effectively.</p><p>This project explores the use of data analytics and machine learning to address credit card fraud detection. By leveraging SQL for exploratory analysis, Python for predictive modeling, and Power BI for data visualization, the project aims to uncover fraud patterns and develop a robust system for real-time fraud identification.</p><h3>Data Cleaning and Preprocessing</h3><p>The dataset was gotten from <a href="https://www.kaggle.com/datasets/mlg-ulb/creditcardfraud/data">Kaggle</a> and Excel was used to clean the data, checking for nulls, duplicate and outliers and ensure proper data types to every column. This step is crucial in understanding the data and ensuring it is in an excellent state for analysis.</p><h3>Data Analysis and Insights</h3><p>Using SQL for exploratory data analysis, the following key insights were gotten.</p><h3>Fraud Cases and Non-Fraud Cases</h3><p>· Fraudulent cases comprised of only 0.173% of the total transactions, which indicates a significant class imbalance.</p><p>· Non-fraudulent transactions dominated the data, which indicates a need for an advanced handling during prediction analysis.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/810/1*7x9OXO3-TgYRg6fcbCidTg.png" /></figure><h3>Transaction Amount Analysis</h3><p>· Fraudulent transactions had an average transaction amount 58% higher than non-fraudulent transactions.</p><p>· This could indicate that larger transaction amount is prone to frauds, making the fraudulent cases financially significant despite their low frequency.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/718/1*1ZKluw2PJW7e0CnZLdCsRg.png" /></figure><h3>Prediction Analysis</h3><p>The class imbalance in the dataset required that certain steps were taken to ensure accurate fraud prediction. Python was utilized in this prediction analysis</p><h3>Anomaly detection</h3><p>· Anomaly detection techniques was performed to identify rare and unusual transaction in the dataset. This helped to isolate the potential fraud indicators in the dataset.</p><p>· Isolating potential fraud indicators helped to reduce noise, highlight suspicious data and making it easier to analyze and make predictions.</p><h3>Synthetic Data Generation</h3><p>· Due to the class imbalance between fraudulent and non-fraudulent cases, synthetic fraudulent classes were generated using oversampling techniques such as Synthetic Minority Oversampling Technique (SMOTE).</p><p>· After the use of SMOTE to balance the dataset, the predictive model could then effectively learn from both classes.</p><h3>Classification Model Training</h3><p>· A classification model was trained using algorithms such as Logistic Regression and Random Forest.</p><p>· The model was evaluated using metrics such as ROC-AUC score, precision, recall and accuracy to ensure reliability of identifying fraudulent cases.</p><h3>Model Results</h3><p>Two machine learning models were trained and evaluated.</p><p>a. Random Forest Classifier:</p><p>· Accuracy: 1.00</p><p>· ROC-AUC Score: 0.9345</p><p>· It achieved perfect accuracy, showing its ability in distinguishing between fraudulent and non-fraudulent transactions.</p><p>· Its ROC-AUC score shows that the model has a 93.45% chance of correctly distinguishing between a random fraudulent transaction and a non-fraudulent transaction.</p><p>b. Logistic Regression:</p><p>· Accuracy: 0.97</p><p>· ROC-AUC Score: 0.9345</p><p>· It has a slightly lower accuracy than the Random Forest but matched in terms of ROC-AUC score, indicating that it can effectively distinguish between both cases.</p><h3>Model Deployment</h3><p>· The two trained models were saved using Joblib, ensuring that they are ready for deployment in real-time fraud detection system.</p><h3>Data Visualization</h3><p>The insights such as class distribution, fraud transaction over time and amount and also transaction analysis was visualized on a dashboard Using Power BI.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1006/1*9B2CIhLG-vSDW_e2O3ApmQ.png" /></figure><h3>Recommendations</h3><p>· Transaction limits should be established and also enhanced customer alerts for notifications of high-risk activities should be developed. This would help prevent fraud before it occurs.</p><p>· Informing customers about proactive fraud prevention measures and also training customer support staff on how to effectively handle queries of flagged transactions.</p><p>· Additional features like geographical location, device information would enhance data quality.</p><h3>Conclusion</h3><p>This project provides a comprehensive approach to analyzing and predicting credit card fraud. The analysis shows the low frequency of fraudulent cases (0.173%) but emphasizes their financial significance due to higher average transaction amounts. The predictive analysis addressed the imbalance issue which enhanced the performance of the classification models.</p><p>By implementing the recommendations, organizations can improve their fraud prevention strategies, enhance customer trust, and minimize financial losses.</p><p>The entire project can be viewed on <a href="https://github.com/Amina-py/Credit-Card-Fraud-Detection">GitHub</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3ba9cd865310" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[LENDING CLUB LOAN DEFAULT ANALYSIS AND PREDICTION]]></title>
            <link>https://medium.com/@akintundetemiloluwa.a/lending-club-loan-default-analysis-and-prediction-1dd5132b82be?source=rss-cc2a9b4c51bb------2</link>
            <guid isPermaLink="false">https://medium.com/p/1dd5132b82be</guid>
            <category><![CDATA[sql-server]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[lending-club]]></category>
            <dc:creator><![CDATA[Temiloluwa Akintunde]]></dc:creator>
            <pubDate>Wed, 15 Jan 2025 12:41:26 GMT</pubDate>
            <atom:updated>2025-01-15T12:41:26.807Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/474/1*b3Ukf1OI-Paqr72Jaycqpg.jpeg" /></figure><h3>Introduction</h3><p>In today’s competitive financial landscape, lending institutions face a critical challenge: managing loan defaults while maintaining profitability. The ability to identify high-risk borrowers and predict potential defaults is essential for minimizing losses and optimizing lending strategies. This article presents a comprehensive analysis of the Lending Club loan dataset, leveraging data analytics and machine learning to uncover insights into borrower behavior and loan performance.</p><p>By utilizing a combination of SQL, Python, and Power BI, this project explores the key factors contributing to loan defaults, builds predictive models to assess default risks, and visualizes findings through dynamic dashboards. The analysis not only highlights trends in borrower demographics, loan attributes, and seasonal patterns but also demonstrates the power of predictive analytics in transforming raw data into actionable insights.</p><h3>SQL Analysis Insights</h3><p>This exploratory data analysis was carried out to uncover patterns and relationship that could enhance loan performance and risk assessment. The key insights derived from the exploration include:</p><p>1. Charged-Off Loans and Fully Paid Loans</p><p>The analysis showed that the charged-off loans accounted for 19.74% of the total loans, which is less than the fully paid loans. This indicates that a significant amount of loans is successfully repaid, though the default rate still poses concern for risk management.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/444/1*tqlyvromf_8SOyZF0XShkg.png" /></figure><p>2. High Risk Borrowers Profiles</p><p>Advanced SQL queries identified the borrower characteristics that has higher risk default rate. Borrowers with the following profile showed the highest risk:</p><p>· Employment Titles: Manager, Supervisor, Teacher and Owner were associated with higher default risk.</p><p>· Employment Length: Borrowers with over 10 years of employment showed a higher likelihood of default.</p><p>· Home Ownership: borrowers with mortgage and rent were more likely to default.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/934/1*UpxrA8bmW2TmOivPeSsPPA.png" /></figure><p>3. Loan Term and Amount Impact on Default Rate</p><p>· A loan term of 36-month with low to moderate loan amount posed the highest default rate, this shows that shorter loan terms and low loan amount has increased risk default.</p><p>· A 60-month loan term with moderate to high loan amount was associated with a higher default rate, suggesting that longer loan terms and larger loan amount increases the risk of default.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/608/1*424XMRmiGeU3dDrMV0zR7A.png" /></figure><p>4. Loan Purpose and Default Rates</p><p>The top three loan purposes with the highest default rates across all grades were:</p><p>· Debt Consolidation: this purpose had the highest default rate; this indicates that borrowers with this purpose may face financial difficulties.</p><p>· Credit Card: borrowers seeking loan for this purpose also has an increased chance of default.</p><p>· Home Improvement: Loans taken for home improvement has a significant default rate across the various grades.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/830/1*683eagX2fIUEipH0ZnMdSw.png" /></figure><p>5. Interest Rates of Charged Off vs. Fully Paid Loans</p><p>The average interest rate for charged off loans was 54.8% significantly higher than the average interest of fully paid loans. This suggests that higher interest rates could be a contributing factor to loan defaults.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/686/1*gihppDLiQKgNqny-aSCYZQ.png" /></figure><p>6. Monthly Trends in Loan Defaults</p><p>· October and July recorded the highest number of charged-off loans (7,264 and 7,229 respectively with default rates of 19% and 20%.</p><p>· December had the lowest number of charged-off loans (4,085) and a default rate of 17%, indicating a seasonal variation in loan performance.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/832/1*sVDr_Y2GcTuz44ULIt7tOQ.png" /></figure><p>7. Impact of Open Credit Lines on Default Rates</p><p>Borrowers with higher credit lines were found to default less than those with low credit lines. This insight indicates that borrowers with better access to credit may be more financially stable and less likely to default.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/712/1*0md78waUuukmTzlUoD-rSg.png" /></figure><p>8. Public Records and Default Rates</p><p>Borrowers with no or low public records were more likely to default than those with high public records. This suggest that a lack or negative financial history may be indicative of higher risk.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/723/1*4FCSS3LxoP08YPEf0w9-hA.png" /></figure><h3>Python Prediction Analysis</h3><p>1. Feature Engineering</p><p>New features were created, and existing features were transformed to improve the model’s predictive power.</p><p>2. Data Visualization</p><p>The relationships between the variables and the target variable (loan status) were visualized for better understanding. This process helped in identifying the most relevant features for prediction.</p><p>· Categorical variables: Plots were used to show the distribution of categorial variables against the target variables to identify trends and patterns.</p><p>· Numerical Variables: A correlation heatmap was used to visualize the relationship between the numerical features and their correlation with the target variable.</p><p>3. Feature Selection</p><p>A selection of relevant columns was made for the prediction model. It focused on choosing the features that had the most impact on the target variable.</p><p>4. Column Transformer and Preprocessing</p><p>A column transformer was applied to combine the preprocessing steps such as encoding categorical variables and scaling numerical variables. This was important because it ensured that the variables were in a well processed state for the model.</p><p>5. Model Building</p><p>A voting classifier was constructed by combining multiple machine learning models such as Random Forest, XGBoost and Logistic Regression.</p><p>The voting classifier would aggregate the predictions from these models to enhance the overall accuracy of the prediction.</p><p>6. Pipeline Creation and Evaluation</p><p>The pipeline was used to streamline the process, ensuring that all preprocessing and model-building steps were applied consistently during the training and testing.</p><p>7. Model Evaluation</p><p>After training the model, its performance was evaluated using the accuracy metrics. The model achieved an accuracy of 80%, indicating a strong ability to predict loan default status.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/970/1*DPym-eYm8LKOfQGB3n6oMQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/522/1*wRdbN_WRrfU8fbKIGbvrpA.png" /></figure><h3>Data Visualization</h3><p>Power BI was used to create interactive dashboards on the dataset overview, borrowers Profile and risk analysis, showcasing the insights from SQL and Python.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1003/1*HnPIqSU5CtS2F-NE61Gjmg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1002/1*jDyeX6ttQ0XPvVRQ6u5wXw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1008/1*5sKJYwXLEY2JY6cGa7YqcA.png" /></figure><h3>Recommendations</h3><p>1. Stricter evaluation criteria should be imposed on borrowers with high-risk characteristics, and loan purposes with historical high default rates.</p><p>2. Lending policies for the various loan terms should be adjusted and interest rates polices should be reassessed to ensure that they aren’t difficult for borrowers. This would help keep a balance between profitability and borrower affordability.</p><p>3. Seasonal default patterns should be monitored. Proactive borrower engagement on months with high default rate should be practiced during these periods. Months with lower default rate encourage growth in business, so lending during this period should be well promoted.</p><h3>Conclusion</h3><p>This analysis was able to provide valuable insights on the lending club dataset and by implementing these recommendations and taking actionable decisions in line with the insight derived can help reduce loan defaults, improve business performance, and enhance borrower satisfaction.</p><p>The entire project can be viewed on <a href="https://github.com/Amina-py/Lending-Club-Loan-Analysis">Github</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1dd5132b82be" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[BANKING CHURN ANALYSIS]]></title>
            <link>https://medium.com/@akintundetemiloluwa.a/banking-churn-analysis-1262d2a7aaac?source=rss-cc2a9b4c51bb------2</link>
            <guid isPermaLink="false">https://medium.com/p/1262d2a7aaac</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[data-visualization]]></category>
            <category><![CDATA[python]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Temiloluwa Akintunde]]></dc:creator>
            <pubDate>Wed, 15 Jan 2025 12:40:15 GMT</pubDate>
            <atom:updated>2025-01-15T12:40:15.385Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ey8W7mKNilNhOeYgPOS5VA.png" /></figure><h3>Introduction</h3><p>Customer churn is a critical challenge for businesses, particularly in the banking sector, where retaining customers is vital for long-term profitability and sustainability. Churn not only impacts revenue but also increases acquisition costs as businesses strive to replace lost customers. Understanding the factors influencing churn and predicting at-risk customers are essential for designing effective retention strategies. This project focuses on analyzing customer churn in a banking dataset using SQL, Python, and Power BI.</p><p>The analysis aims to:</p><ul><li>Identify patterns and trends that influence customer churn, such as demographics, balances, credit scores, and tenure.</li><li>Predict churn likelihood using machine learning models to enable proactive retention efforts.</li><li>Visualize insights through dashboards, providing actionable information for stakeholders.</li></ul><h3>SQL Insights</h3><p>1. Overall Churn Rate</p><p>20.37% of customers have churned, while the majority (79.63%) remain active. This indicates that the bank retains most of its customers but understanding the reason of the 20.37% churn is needed to improve retention.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/648/1*fN2-UE-NwRnXf3r4XsFzag.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/648/1*fN2-UE-NwRnXf3r4XsFzag.png" /></figure><p>2. Customer Demographics and Churn</p><p>· Geography: Customers from all 3 countries exhibit churn, but females have the highest churn rate across the geographies.</p><p>· Gender: Females churned more than males across all age groups.</p><p>· Age Group: Adults aged 31–55 has the highest churn rates among both males and females.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/872/1*l549q5rCxme3qb0iX7Na0w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/888/1*t163lNadfLYLyPDRr3jP7Q.png" /></figure><p>3. Salaries and balance influence on churn rate</p><p>· Customers with high balances regardless of their salaries had the highest churn rate.</p><p>· A significant number of customers with no balance but high salaries also churned, as did those with low balances and high salaries.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/738/1*g57BzOV44E5hfEaVyoCqww.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/290/1*UTV8hOUkEsv1QdI8sjvPLw.png" /></figure><p>4. Credit scores</p><p>· Customers with credit scores in the range of 300–739 exhibited a high churn rate. While those with excellent credit score (749–799) had the lowest churn rate.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/576/1*KmGsjzL5ASPx0OaDWxvZeg.png" /></figure><p>5. Tenure and Churn</p><p>· New Customers (Tenure 0–2 years): Lower churn rate.</p><p>· Old Customers (Tenure 3+ years): higher churn rate by 74% compared to new customers</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/577/1*y3_j5c55D3wPYrFmbOuxcA.png" /></figure><p>6. Bank Activity and Churn</p><p>· Customers with only one bank product had a significantly higher churn rate compared to those with multiple products.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/974/1*1OWQ_tIN92TrbGNvypYnhQ.png" /></figure><h3>Prediction Analysis</h3><p>A prediction model was developed using Machine Learning to predict customer churn and identify patterns influencing churn.</p><p>1. Feature engineering</p><p>Categorical variables were plotted against the target variable to identify patterns. Numerical variables were analyzed for their correlation with churn.</p><p>2. Data Preprocessing</p><p>Column transformer was implemented to streamline preprocessing of standard scaling and one-hot encoding.</p><p>3. Modeling Process</p><p>Multiple machine learning models were tested to ensure robustness and accuracy such as Logistic Regression, Random Forest Classifier, Decision Tree Classifier, Support Vector Machine, K-Nearest Neighbors, and XGBoost Classifier.</p><p>A pipeline was built to combine the data preprocessing and model training, ensuring a streamlined and repeated process.</p><p>4. Model Evaluation</p><p>The models were evaluated using accuracy and mean cross-validation scores.</p><p>The Random Forest model showed the highest accuracy (86.34%) and cross-validation scores (0.86), making it the most reliable for predicting churn and the top performing model among the models tested.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/756/1*3NJcdLxHzu6wq0VY_zQGpw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/430/1*G2TxGvZlGZMd9whcOveplQ.png" /></figure><h3>Visualization</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1004/1*onMTnKtuN0ppSRoGsNe3FQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/998/1*k6gJ57sdOIY-yffa0AGWEg.png" /></figure><h3>Recommendation</h3><p>1. Customer retention strategies should focus more on developing personalized incentives, loyalty campaigns or proactive engagement that would encourage female in the 31–55 age group across all geographies and also other customers.</p><p>2. Exclusive benefits like better interst rates, or premium services to retain high balance customers.</p><p>3. Financial counselling or proactive credit building engagement can be created to support customers in improving their credit scores. This would make banking more accessible for such customers.</p><p>4. Regular feedback from long term customers to understand their issues would enable identify and solving such pain points easier for customer service officials. This would make the customers relaxed and more confident in the banking operations.</p><h3>Conclusion</h3><p>This analysis has provided actionable insights into customer behavior and churn patterns. The insights and predictions showed the importance of personalized engagement, proactive retention effort and creation of product that meets a variety of customers there by reducing churn and improving customer loyalty.</p><p>The entire project could be viewed on <a href="https://github.com/Amina-py/Banking-Customer-Churn">Github</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1262d2a7aaac" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Heart Health Analysis and Prediction]]></title>
            <link>https://medium.com/@akintundetemiloluwa.a/heart-health-analysis-and-prediction-f58ae5c01992?source=rss-cc2a9b4c51bb------2</link>
            <guid isPermaLink="false">https://medium.com/p/f58ae5c01992</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Temiloluwa Akintunde]]></dc:creator>
            <pubDate>Sat, 06 Apr 2024 17:14:04 GMT</pubDate>
            <atom:updated>2024-04-07T09:40:08.934Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/952/1*pjeqfTHncCrDeRLnqzTjWw.jpeg" /><figcaption>Source: <a href="https://images.app.goo.gl/gwkX8mN8cwJ33fzS8">Google</a></figcaption></figure><p><strong>INTRODUCTION</strong><br>Maintaining a healthy lifestyle and giving priority to heart health are essential for preventing heart diseases. Heart health is very crucial as it impacts both the physical and mental well-being of a person. <br>The heart health stats dataset comprises of demographic information, vital signs, and heart status indicators for individuals. It includes features such as age, gender, blood pressure, cholesterol levels, and whether the individual has experienced a heart attack.</p><p>This dataset was useful in analyzing the trends relating to cardiovascular health and the risk factors associated with heart disease. The dataset can be gotten from <a href="https://www.kaggle.com/datasets/mahad049/heart-health-stats-dataset">Kaggle</a>. The tools used in the project are Python and SQL.</p><p><strong>AIM </strong><br>The aim of this project is to identify patterns and formulate personalized recommendations for heart disease prevention and management.</p><p><strong>DATA CLEANING</strong> <br>Python was used to clean and prepare the data. Necessary libraries were first imported, and pandas was used to import the dataset.</p><p>The top 5 rows and shape of the dataset were checked using the head function and shape method.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/1*1rtn9EFzcgzXu7hSoBy1vA.png" /></figure><p>The presence of null values was then checked, and it was observed that there were no null value present in the dataset. <br>Presence of duplicate values were also checked, and there were 535 duplicates present. The duplicates were dropped to carryout proper analysis.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*49WMn5CCtIjK6CsnU2bwtA.png" /></figure><p>The shape of the dataset as checked again to understand the amount of data available to work on. The blood pressure column was separated into two columns and the data type of the columns were changed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*K_wlpOKnN-g5DWmDxk_v7w.png" /></figure><p>Information on the dataset was checked to check the non-null count, dtype of each column and memory usage of the dataset. The cleaned data was then saved using the pandas “to_csv() method”.</p><p><strong>DATA ANALYSIS</strong></p><p>The cleaned dataset as then imported into SQL database to extract insights from the data.<br>1. Which gender has the most numbers of patients with heart attack?</p><p><em>(0-no heart attack &amp; 1-has heart attack)</em><br>The result shows that males are more prone to having heart attack than females. This could be due to the difference in hormones, cholesterol build-up, physical activities carried out by the respective genders.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/603/1*jQJx0gAMmyHc866Vs2mXew.png" /></figure><p>2. Which age group is more susceptible to heart attack?</p><p>The age group column was created to group various ages and the query result showed that people above the ages of 45 are highly susceptible to heart attack. This proves that as people age the rate at which they are susceptible to heart diseases increases.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/638/1*v6fcuuAQ3vvV-viYa_1SCQ.png" /></figure><p>People aged above 45 has a higher chance of attaining heart attack. <br>3. How does the BMI influence the chances of having heart attack?</p><p><em>BMI — Body Mass Index</em></p><p>BMI is an index of eight-for-height and it is used to classify various body weights. The dataset has people between the ranges of normal weight(18.5–24.9) and overweight(25.0–29.9). Result of the query shows that people having BMI of 25–26 had the highest count of having heart attack while those with BMI of 24–25 had the highest count of not having heart attack.</p><p>This proves the fact that excess weight can lead to the accumulation of cholesterol or fatty material in the blood vessels which can can get the heart damaged and lead to heart attack.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/662/1*jZbwpWP_56lOtNcABeBJBw.png" /></figure><p>4. How does smoking influences heart attacks?</p><p>The result of the query shows that smoking does not show a distinct effect on whether it could lead to heart attacks or not. So the possibility of smoking leading to heart attack could be dependent on other factors.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/754/1*PFsBbL9tDegeFzZeV4ogfA.png" /></figure><p>5. Is there a correlation between cholesterol level, glucose level and blood pressure on having heart attack?</p><p>From the map it can be seen that cholesterol level, glucose level and blood pressure(Systolic &amp; Diastolic pressure) has a high positive correlation with heart attack. This means that the higher those pressures are the higher the rate of having heart attack.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/865/1*M5hcc5EEBiX_mMOJkRmHTw.png" /></figure><p>HEART HEALTH PREDICTION</p><p>A simple machine learning model using Python library “Scikit-learn” was built to predict heart attack risk based on some features.</p><p>Necessary libraries were imported and Logistic Regression (Binary classification) was used to build the model. Then the dataset set was split into train and test set using the train_test_split method .</p><p>The train set was used to train the model and the test set was used in predicting and ensuring that the model works well when new information is used on it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*um15W6es-LFFUAMRrd7ehw.png" /></figure><p>Then StandardScaler was used to remove the mean and scaling to unit variance. The logistic regression model was built and then fit to the train set.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*b-HKNoHRXtlfVd9fD0pVNw.png" /></figure><p>The train set was then used for prediction and it gave a r2 score of 0.9 which shows that the model worked on the training set. The test set (new data) was used for prediction and gave a r22 score of 1.0 showing that the model works perfectly for unknown data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1020/1*2V_s4YhWpWw-DyAqSjmjFw.png" /></figure><p>RECOMMENDATIONS</p><ol><li>Everyone should be encouraged to adopt a healthy lifestyle by incorporating regular exercise, stress management methods, balanced diet, maintain healthy weights and avoiding smoking and excessive alcohol consumption.</li><li>Regular health check-up and monitoring of cholesterol level, blood pressure level, glucose level should be of high importance especially for people aged above 45 years.</li><li>Creating awareness and educating individuals about risks of heart diseases and also preventive measures that can help to reduce the risks. Medical outreach in various communities, creation of support groups, affordable healthcare can also help in reducing risk of heart attacks.</li></ol><p>CONCLUSION</p><p>This project was able to help understand the effect of various factors on heart health and the recommendations aims in encouraging individuals to take proactive steps in improving their heart health. The model would also serve as a guide to predict the chances of an individual risk of having heart attack.</p><p>This project codes can be viewed in <a href="https://github.com/Amina-py/Python-projects/tree/main/Heart%20Health">github</a>.</p><p>Thank you for reading my article!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f58ae5c01992" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[ADIDAS SALES ANALYSIS]]></title>
            <link>https://medium.com/@akintundetemiloluwa.a/adidas-sales-analysis-258286a671ec?source=rss-cc2a9b4c51bb------2</link>
            <guid isPermaLink="false">https://medium.com/p/258286a671ec</guid>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[adidas]]></category>
            <category><![CDATA[python]]></category>
            <category><![CDATA[power-bi]]></category>
            <dc:creator><![CDATA[Temiloluwa Akintunde]]></dc:creator>
            <pubDate>Sat, 17 Feb 2024 23:01:33 GMT</pubDate>
            <atom:updated>2024-02-17T23:01:33.883Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/736/1*JPHU8IYXa-xl86bTWqbQdA.jpeg" /></figure><p><strong>INTRODUCTION</strong></p><p>Adidas is one of the largest manufacturers of sports shoes and clothing. This sales analysis of the Adidas Sales dataset helps to give better insights to trends and details in the data such as retailer, location of the sale, total sales revenue, type of product sold, and any other relevant information. This dataset was gotten from <a href="https://www.kaggle.com/datasets/heemalichaudhari/adidas-sales-dataset">Kaggle</a> and analyzed using Python and visualized with Power BI.</p><p><strong>AIM</strong></p><p>Analyzing the adidas sales dataset to extract meaningful insights to understand the sales trend and improve decision making on the market strategies to be implemented.</p><p><strong>DATA CLEANING</strong></p><p>Python was the tool used to clean and prepare the dataset, necessary libraries like pandas, seaborn, matplotlib were imported to carry out proper analysis on the dataset.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6BQqNooPS4yWf1PpX-Qxqw.png" /><figcaption>Importing Libraries and the Dataset</figcaption></figure><p>The top 5 rows was checked to observe the dataset and an unnecessary column was dropped.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*IIhdzv_h7cqW1PFAezajfw.png" /><figcaption>Top 5 Rows</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mnC-EVjWizAl2-2VZETHKA.png" /><figcaption>Column Check</figcaption></figure><p>Checking the number of rows and columns using the shape method.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gNteNab4K4KoLvT89-f_sA.png" /><figcaption>Shape of the Dataset</figcaption></figure><p>Presence of duplicate values was checked.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a49Fp8Ngyo8Zy_45WCa6Jg.png" /><figcaption>Duplicate Check</figcaption></figure><p>Informations such as column name, non-null values count, dtypes, memory usage was checked using info() function</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0iMJKSYro-TJwOzfK6sA7A.png" /><figcaption>Information on the Dataset</figcaption></figure><p><strong>Exploratory Data Analysis (EDA)</strong></p><p>EDA is used to analyze, investigate and summarize the content in a dataset by showing their main characteristics, discover patterns, spot anomalies, test a hypothesis, or check assumptions and it often employ data visualization methods.</p><p>Univariate analysis was used to understand the content of some columns with the aid of tables, charts, boxplot and histograms.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/831/1*sMi67SFYaVy_Pgfg2CRvYQ.png" /><figcaption>Bar plot of Region Column</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/855/1*Rc-ogH5lXFRTz_jzQFQ5UQ.png" /><figcaption>Bar plot of Month column</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/853/1*FaAbTadD3PoKj_6Pv4ZpiQ.png" /><figcaption>Bar plot of Retailer column</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7BNTxeSYFRvCipYEiXBYNQ.png" /><figcaption>Table showing unique values in product column</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1006/1*N523kyVDf8RWKQ4b_gNeKA.png" /><figcaption>Box plot showing price distribution</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1008/1*ierc-666VJvzRgr4zgOTeA.png" /><figcaption>Histogram plot of Unit Sold Distribution</figcaption></figure><p><strong>Insights</strong></p><ol><li>Sales Monthly Trend</li></ol><p>January had the highest number of sales and this could be factored toward people wanting to start the year with new wears. Men’s Street Footwear and Women’s Apparel were the most sold product in the month.</p><p>There was a huge decrease in the sales in June and from further analysis South and Southeast had the least number of sales made in June having Men and Women Apparels as the least products sold.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/886/1*reF3Ku1h8p-cQk7IC0AgSA.png" /><figcaption>Monthly Trend of Sales</figcaption></figure><p>From the plot, it can be seen that there is no consistent trend in the sales made monthly.</p><p>2. Retailers Analysis</p><p>West Gear and Foot Locker contributed the highest to the sales by $242,964,261 and $220,094,606 respectively. Most of their sales were made through In-store sales method having Men’s Street Footwear and Women’s Apparel as their top products.</p><p>Amazon made the least sales of $77,698,884 and most of it’s sale was made through the Online sales method.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/735/1*TpOGTFdj_ZDgJmOvqRSvRg.png" /><figcaption>Total sales for each product by retailer</figcaption></figure><p>3. Region Analysis</p><p>West region made the highest profit when compared with other regions by generating $89,608,607 profit from the sales having San Francisco as its top performing city earning a profit of $10,261,188 and its least performing city is Salt Lake City gaining a profit of $3,873,353.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1008/1*4Y8M2eKQoL2CbFsvVFyrsw.png" /><figcaption>Top Cities in each Region</figcaption></figure><p>Midwest region made the lowest profit of $52,810,769. It has its top and least performing city as Detroit and Omaha generating profit of $8135844 and $2439433 respectively.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1007/1*l2-0QIHjopRxsGJ7-GL8Wg.png" /><figcaption>Bottom Cities in each Region</figcaption></figure><p>4. Correlation Analysis</p><p>The correlation between number of units sold and operating profit is strongly positive (0.892), this means that the number of units sold strongly influences the profit. The higher the number of units sold per products the operating profit increases.</p><p>The correlation between the price per unit of each product and operating profit is weakly positive (0.393), this means that the price of the product slightly influences the operating profit.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5UBtFOMU8nwOjSk-UKwM_Q.png" /><figcaption>Correlation Analysis</figcaption></figure><p>5. Sales Method Analysis</p><p>· Frequently used sales method</p><p>The Online sales method has a count of 50.7% making it the most preferred sales method for customers. This could be linked to the ease of ordering for the product from online retailers.</p><p>Outlet and In-store sales method has counts of 31.3% and 18% respectively.</p><p>· Sales method with highest total sales and profit</p><p>In-store sales method has the highest sale of $3,566,435,665 and profit of $127,590,973.</p><p>Outlet sales method has a total sale of $29,558,547 and profit of $107,987,205.</p><p>Online sales method has the lowest total sale of $24,767,811 and profit of $9,553,370.</p><p>The In-store sales wasn’t used frequently used but it generated the highest revenue, this shows that customers who purchases using this method buy products in larger quantities than those who used the Online sales method.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/958/1*q4ezvOLAwtOe2ClBo2sioQ.png" /><figcaption>Sales method analysis</figcaption></figure><p><strong>Data Visualization</strong></p><p>This interactive dashboard was created using Power BI and it can be assessed <a href="https://app.powerbi.com/view?r=eyJrIjoiMjAxMzQ1NDQtOTEzZC00YjZjLThmODAtNTdlZjIzZGFlMDVmIiwidCI6IjFhOGM3NmI4LWUxNTMtNGExNi1hNGExLTliNzc2ZTgyN2Y0NiJ9">here</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*k3xS1dVF3BocDQJLyZLNNw.png" /><figcaption>Adidas Sales Dashboard</figcaption></figure><p><strong>Recommendations</strong></p><ul><li>Creation of seasonal promotion post or advert to boost sales in months that generated low revenue.</li><li>Partnership with popular companies that could help boost sales in specific regions or cities; this could attract more customers that are loyal to such company.</li><li>Improving sales facilities for the various sales method; online sales method was popular but generated low revenue. Better online sales method could be designed to enable sales purchase easier, safer and linked with benefits such as discounts for customers.</li></ul><p><strong>Conclusion</strong></p><p>This sales analysis has been able to extract valuable insights from the Adidas Sales dataset and provide recommendation that could help improve the sales and make valuable decisions.</p><p>Thank you for reading my article!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=258286a671ec" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[HOTEL BOOKING ANALYSIS]]></title>
            <link>https://medium.com/@akintundetemiloluwa.a/hotel-booking-analysis-37235d0c47f3?source=rss-cc2a9b4c51bb------2</link>
            <guid isPermaLink="false">https://medium.com/p/37235d0c47f3</guid>
            <category><![CDATA[excel]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[data-visualization]]></category>
            <dc:creator><![CDATA[Temiloluwa Akintunde]]></dc:creator>
            <pubDate>Wed, 17 Jan 2024 22:47:08 GMT</pubDate>
            <atom:updated>2024-01-17T22:47:08.696Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/736/1*7Y1MuKphiu8c1kSpRO1ZWg.jpeg" /><figcaption>Source: Pinterest</figcaption></figure><p>INTRODUCTION</p><p>Hotel booking analysis is important for every hotel business to examine trends, patterns and insights that can help make better business decisions, track revenues and marketing effectiveness. This project enabled me to further practice and sharpen my excel skills extensively by identifying the key metrics, extracting meaningful insights and providing recommendation.</p><p>AIM</p><p>To analyze the hotel bookings dataset and extract actionable insights that can optimize the company’s services for international travelers.</p><p>DATA SOURCE</p><p>This dataset was gotten from <a href="https://twitter.com/datafrik_co/status/1744306719453290500?t=_Uw0O_O45CSMFjHVBBOHTA&amp;s=19">DatAfrik</a> data challenge comprising of hotel bookings between January 1st 2010 and December 31st 2019, from a company who helps international travelers around the world secure hotels in destination countries.</p><p>DATA CLEANING AND PREPARATION</p><p>1. Loaded the dataset in Excel and ensured that each column had right data types.</p><p>2. Removed duplicates from the data ribbon and checked for the presence of null values. Then I checked for inconsistent spellings in each columns using filter.</p><p>3. New columns were created:</p><p>· No of days using the formula “Check-Out Date — Check-in Date”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HUlRBldhFXUiRBN2gStrBA.png" /><figcaption>Formula for deriving &quot;No of Days&quot; spent in booked hotel</figcaption></figure><p>· Age Group using the IF formula.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*36pNIOtw75OsVuf_u6fL7w.png" /><figcaption>IF statement formula for grouping various ages</figcaption></figure><p>DATA VISUALIZATION AND ANALYSIS</p><p>The dataset was visualized with excel to create an interactive dashboard to identify the trends and patterns present.</p><p>Statistical analysis using Stata was used to make inferences using the chi square test, correlation test and Anova test to check the relationship between various variables in the dataset.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5zaiLwLjaPFwMUIqIUrw5w.jpeg" /><figcaption>Hotel Booking Dashboard</figcaption></figure><p>INSIGHTS DERIVED</p><ul><li>The company had 66535 bookings from 2010–2019, along the years January and December had relatively high number of bookings this could be due to the festive periods and holidays.</li><li>Majority of the customers are from Thailand, Malaysia, Singapore, and Indonesia. Generally, the female customers are slightly more than the male customers. Internet Banking is the most used method of payment by the customers.</li><li>Among the age groups, middle aged adult (31–45 years) made the most bookings while young adults (18–30 years) had the least number of bookings. The average number of days spent in the hotel by all age groups is 3 days.</li><li>There is no relationship between the origin country and choice of destination country so, the choice of which country to travel to is dependent on the customer. This relationship was determine by analyzing the relationship between the origin country and destination using chi square test. Although New Zealand, Nepal and Egypt are the top 3 most booked countries to visit.</li><li>Most customers book hotels having high ratings such as Enchanted Isle, Malibu Motel, Ace Hotel. From the Anova analysis it showed that hotel rating has 34.39% impact on the booking price. Therefore, there is a slight relationship between hotel rating and its booking price.</li></ul><p>RECOMMENDATIONS</p><p>Based on the insights and findings from the analysis, these recommendations were made:</p><ul><li>Focus marketing strategy to countries in other continent by creating media outlet such as Instagram posts or blogs on the services the company renders, this could improve the company’s growth and revenue.</li><li>Creating stronger partnership with hotels that provides special packages and also with high and moderate ratings so that customers would be assured of their comfortability.</li><li>Creation of a detailed brochure of best hotels that would suite different targeted customers. This could help in influencing the customer decision of hotel or country to visit.</li></ul><p>CONCLUSION</p><p>This analysis has provided a brief description, valuable insights and recommendations of the company’s hotel bookings. The suggested recommendation could help the increase the company’s growth, revenue and customers satisfaction.</p><p>Thank you for reading!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=37235d0c47f3" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>