Hybrid Movie Recommandation System by description and category

Recommend favorite movies based on user input description and favorite movie category

Author

Howard W. Chung

Introduction

This library implements Movie Recommandation website by using sentence embedding model BERT & text feature extraction method TF-IDF, to select the movies that users might be interested in. We build this system with two flask model for Predict Vector ModelEnd and Website WebEnd.

Requirement

• python>=3
• flask>=1.0
• numpy>=1.15
• pandas=0.24.2
• bert-serving-client=1.10.0
• requests=2.21
• json=0.9.2
• pickle=0.7.5
• os

Start Up

Start Flask of Model End

1. Setup output port in line 20 in RecommandMovie.py
   
2. Start ModelEnd Flask
   

$ python3 RecommandMovie.py

Deploy BERT Serving client

3. Download BERT base model
   
4. Start BERT Serving client
   

$ PTHNAME="./uncased_L-12_H-768_A-12" #Path of Model
$ bert-serving-start -model_dir ${PTHNAME} -num_worker=1

Start Flask of Website End

5. Setup ModelEnd IP and port in line 71 in app.py
   
6. Start WebEnd Flask
   

$ python3 app.py