Projects about machine learning methods by python3
Built Logistic Regression (LR) and Naive Bayes (NB) models to predict a new bidder is bot or a human. Evaluated models’ performances using Log Loss Evaluation, ROC curve and Precision-recall curve, which indicated the LR model was better than the NB model.
Built Single Decision Tree, Random Forest and Ada-boosting Tree to make prediction. By tracking the Normalized Discounted Cumulative Gain (NDCG), the accuracy rate increased by 11-23%, which could increase user engagement.
Built 3 different Collaborative Filtering models of recommender system using Unsupervised Learning methods.
Centered Cosine Similarity based method, Singular Value Decomposition (SVD) method and Funk SVD method (100 iterations in Gradient Decent process)
Calculated the Root Mean Square Errors (RMSE) and compared the top 10 predicted and actual favorite movies for a certain user. The numbers of correctly predicted were 3, 6 and 7, respectively.
Built Linear Regression, Ridge Regression and Lasso Regression models. Used K-fold Cross Validation to compare the Mean
Absolute Error (MAE) and Root Mean Squared Error (RMSE), which indicated the Lasso model was the best.
Made prediction using Lasso Regression model. Used ranking table and scatter plot to compare the predicted results with the real
results. The accuracy is 70%, and the real champion was predicted the 2nd.