In this assignment, I dove into predicting and managing employee attrition for IBM, utilizing advanced modeling techniques such as Logistic Regression, Random Forest, and Gradient Boosted Trees, among others. My analysis aimed at identifying key predictors of attrition, enabling targeted retention strategies. Through exploratory data analysis and preprocessing, I prepared the dataset, focusing on variables like Education, Job Satisfaction, and Work-Life Balance, crucial for understanding employee experiences. The models were evaluated using accuracy metrics, with a special emphasis on the Maximum Payoff metric to assess the financial impact of attrition predictions. My findings led to actionable recommendations, such as managing overtime effectively and revising compensation packages, aimed at reducing attrition and fostering a more stable workforce.
In this assignment, I applied advanced modeling techniques to predict mortgage delinquencies for Fannie Mae, aiming to enhance risk management practices by identifying potential defaults. Through logistic regression, gradient boosted trees, random forests, and Nystroem Kernel SVM, I analyzed data to pinpoint delinquency indicators. The models were evaluated using accuracy metrics and the Maximum Payoff metric to assess economic impacts. The analysis revealed key predictors of mortgage defaults, including borrower credit score, property state, CLTV, and loan purpose, leading to actionable recommendations for mitigating risk. This predictive endeavor aims to bolster Fannie Mae's lending strategy, ensuring financial stability and contributing to a healthier housing market.
In this project, I focused on predicting and understanding customer churn for a telecom company using advanced analytics. The RandomForest Classifier proved to be the most effective model based on its financial payoff, highlighting the importance of customer service interactions, international plan subscriptions, daily usage, and charges as key churn predictors. Recommendations included enhancing customer service, revising pricing and plans, and improving billing transparency to reduce churn and increase customer loyalty, aiming for better retention strategies and financial outcomes.
In this assignment, I worked on predicting which clients are likely to subscribe to term deposits, a crucial part of a bank's marketing strategy. I used logistic regression and decision tree models, drawing on detailed client data to improve predictions. My goal was to find ways to make the bank's marketing efforts more efficient by identifying likely subscribers accurately. This way, the bank could use its resources more effectively and increase the success rate of the campaigns. I evaluated the models based on several performance metrics, including Maximum Profit, to understand the financial impact of marketing strategies better.
In this assignment, I focused on building a logistic regression model to predict personal loan acceptance among bank customers. By examining and pre-processing the Universal Bank dataset, which includes comprehensive customer demographic information, I aimed to determine which customers are likely to accept a loan offer. The model's performance was evaluated based on metrics like Recall, Precision, F1 Score, Error Rate, Accuracy, and ROC AUC, revealing that the model has good predictive value, especially when comparing it to a naive model. Key factors in predicting loan acceptance included Income, Education, CD Account, Family, and Credit Card, with Income being the most significant predictor. The analysis suggests that higher income and education levels, as well as the presence of a CD account, are indicative of a greater likelihood to accept personal loans, aligning with financial behavior insights.
In this assignment, I used DataRobot to build linear regression models to predict the market values of condos in Miami, with an emphasis on features like square footage, number of bedrooms, and bathrooms. The models were evaluated based on metrics like R², MAPE, MAE, and RMSE, but they didn't meet the expected R² level of over 0.9, suggesting that the model's predictive power is not as high as desired. Square footage emerged as the single best predictor of property values, which aligns with the general understanding that property size significantly influences real estate pricing.
This document contains 15 unique writing samples, from media releases, fact sheets, news releases, public service announcements, feature pitches, and more.
Copyright © 2024 Elizabeth Rocks - All Rights Reserved.
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.