Machine Learning Homework Solution Notebooks (UCF CAP5610)

1. Data Preprocessing and Analyze by pivoting features of Titanic dataset

• Titanic dataset: https://www.kaggle.com/c/titanic/data
• Data preprocessing and data analysis

2. Analyze the Titanic data and find whether the individuals from the test dataset had survived or not

Feature Engineering

• Select a set of important features VarianceThreshold Feature Selection using sklearn.feature_selection
• SelectKBest Univariate Feature Selection
• SelectFromModel for Logistic Regression
• Recursive Feature Selection

Build Random Forest and Decision Tree Models

• Build Random Forest and Decision Tree models with the Titanic dataset
• Apply the five-fold cross validation of the Decision Tree and the Random Forest learning algorithms to the Titanic data to extract their average classification accuracies.
• Learn a Decision Tree model with the Titanic data using Gini Index and plot the Decision Tree.

3. Support Vector Machines

• Construct a support vector machine that computes the kernel function.
• Learning SVMs on the Titanic dataset.
• Five-fold cross validation classification accuracies on Titanic training set, with respect to the linear, quadratic, and RBF kernels.
• Tuning the hyperparameters and applying GridSearchCV for each SVM kernels

4. Unsupervised Learning

Football team clustering

• Use Euclidean distance as the distance metric. First, perform one iteration of the K-means algorithm and report the coordinates of the resulting centroids. Second, use K-Means to find two clusters.
• Use Manhattan distance as the distance metric. First, perform one iteration of the K-means algorithm and report the coordinates of the resulting centroids. Second, use K-Means to find two clusters.

K-Means Clustering with Real World Dataset

• Implement the K-means algorithm. K-means algorithm computes the distance of a given data point pair. Replace the distance computation function with Euclidean distance, 1- Cosine similarity, and 1 – the Generalized Jarcard similarity
• Apply different distance matrices.
• Compare the results.

5. Machine Learning for Recommender Systems

• Compute the average MAE and RMSE of the Probabilistic Matrix Factorization (PMF), User based Collaborative Filtering, Item based Collaborative Filtering, under the 5-folds cross-validation
• Compare the average (mean) performances of User-based collaborative filtering, item-based collaborative filtering, PMF with respect to RMSE and MAE. Which ML model is the best in the movie rating data?
• Examine how the cosine, MSD (Mean Squared Difference), and Pearson similarities impact the performances of User based Collaborative Filtering and Item based Collaborative Filtering. Plot your results. Is the impact of the three metrics on User based Collaborative Filtering consistent with the impact of the three metrics on Item based Collaborative Filtering?
• Examine how the number of neighbors impacts the performances of User based Collaborative Filtering and Item based Collaborative Filtering? Plot your results.
• Identify the best number of neighbor (denoted by K) for User/Item based collaborative filtering in terms of RMSE. Is the best K of User based collaborative filtering the same with the best K of Item based collaborative filtering?