Data Science Projects
I was able to work on various projects that attempt to span the entire data science cycle from Business Understanding, Data Understanding, Modeling, Evaluation, and Deployment - through completing my minor in data science and taking the Data Science for Business course in my masters. These projects entail data cleaning and manipulation, exploratory data analysis, the creation of predictive models, and the concise presentation of results in a clear and meaningful manner.
• Cleaned and filtered the loan default data set, conducted k-means and PCA as exploratory data analysis and drew inferences; rectified the limitations of the dataset by oversampling, such as bias towards non-defaulters.
• Developed predictive models using logistic regression with lasso, random forest, gradient boosting and XG boosting and evaluated the best fit using the metric “area under the curve” and accuracy score of the model; Conducted hyper-parameter tuning using python for all four models to determine the optimal combination of estimators for each model.
• Conducted a principal component analysis and k-means cluster analysis to identify relationships between the voter preference and their characteristics using R and Python.
• Engineered a model to predict the winning spread of Obama over Clinton; tested out Linear regression with interaction, Random Forest and Linear regression with Lasso and compared performance using K-fold cross validation (out of sample R2).
• Developed a campaigning strategy for a selected presidential candidate based on whether certain states are predicted to be swing states or not; Analyzed the data to identify which demographic of the population should be targeted for the candidate to win.
• Developed a two-step methodology, a popularity-based model, and a collaborative model based on user ratings to identify and develop the logic to be used for the recommendation system.
• Using user input, a model was created to forecast book recommendations.
• Developed a complete front-end design and used HTML and CSS to build a website.
