Keywords - Neural Networks, Pytorch, EDA, data bias, Sparse AutoEncoders (SAC), Restricted Boltzmann Machine (RBM), Recommendation system
Neural networks (NN) are powerful in predicting new customer preferences based on historic data.NN based recommendation systems have shown to outperform other traditional ML methods.In this notebook, we will use real world data of clothing sales and customers to make predictions for future customer choices.Along with predictions, we will perform Exploratory Data Analysis (EDA) to gain some valuable business insights.
Step 1 - Exploratory Data Analysis (EDA)
Step 2 - Reducing data to tackle class imbalance (CI) and difference in proportion of labels (DPL)
Step 3 - Building Simple feedforward NN to predict future customer preferences
Step 4 - Building Recommendation System using Sparse AutoEncoders (SAE)
Step 5 - Building Recommendation System using Restricted Boltzmann Machine (RBM)
Business Questions
Can we recommend similar clothing items to existing customers based on their age groups and past ratings?
Can we predict which type of clothing a new customer would like and recommend them those to increase our sales?
Are there specific clothing items or classes that receive consistently positive or negative reviews?
What are the most popular clothing classes based on the number of reviews?
Age Group Preferences:
Are there certain age groups that are more likely to rate clothing items positively?
Do certain age groups tend to give higher or lower ratings?
Do different age groups prefer different types of clothing?
Are there certain age groups that consistently recommend specific clothing items?
Is there a correlation between the age of customers and the ratings they provide?
Recommendation Systems
Comments