Topic
Performed feasibility study using WEKA on one of the three key areas machine learning areas identified for the made-up finance company. Submitted a report, with screenshots from the model using WEKA.
Summary
Following the CRISP-DM methodology, created a banking customer churn analysis using a Kaggle dataset. Analysed supervised learning binary classification machine learning models k-nearest neighbor, decision tree, support vector machine and random forest. Reviewed dataset optimisation, hyperparameter tuning, and imbalanced and oversampled datasets. Trained on 90% with 10-fold cross-validation. Reviewed various metrics including AUC-ROC. Random forest outperformed the other models, even with hyperparameter tuning. The number of products and age were the most important parameters.
Tutor Feedback
96% (Distinction) Overall, outstanding work done. Well done for this high quality work. It was a delight to read. There is excellent justification of business context and use of data and algorithms. The implmentation was robust. You demonstrate excellent knowledge and understanding of machine learning modelling with WEKA and the algorithm used to train the model. You also employed good modelling practices by exploring the effect of change in various experimental setups on overall performance and quality of outputs which is excellent. Recommendation: Given the size of the dataset, you could mention the application of Deep Learning algorithms as a potential future work.
