A rolling case study that contains implementation of supervised and unsupervised learning algorithms.

About the dataset:

The Amazon Fine Food Reviews dataset is available on kaggle. The Amazon Fine Food Reviews dataset consists of reviews of fine foods from Amazon. Number of reviews: 568,454 Number of users: 256,059 Number of products: 74,258 Timespan: Oct 1999 - Oct 2012 Number of Attributes/Columns in data: 10

I have done basic EDA but for a detailled EDA please check out this link

Of all the features available, we only considered the ones with textual data (Reviews and for feature engineering Summary) for our predictive modelling. We have used following vectorization techniques to convert text into its vector form:

1. Bag of Words
2. TF-IDF
3. Average word2vec
4. TF-IDF weighted word2vec

Objective:

Given a review, determine whether the review is positive (rating of 4 or 5) or negative (rating of 1 or 2)

The Supervised Learning Algorithms used are:

1. Decision Trees
2. kNN
3. Logistic Regression
4. Navie Bayes
5. Random Forest, Gradient Boosting Decsion Trees, lightGBM
6. Support Vector Machine

The Unsupervised Algorithms used are:

1. k-Means
2. Hierarchical Clustering
3. DBSCAN
4. Truncated SVD
5. Dimensionality reduction technique: t-sne