Movie recommender system and algorithm comparison
SID: 56641800 Name: Du Junye
Python Verison and Package version:
Note: Please make sure that the packages are installed correctly to make the program run normally.
Python version and cuda configuration:
- Python 3.9.7
- cuda 11.0
Missing value handling and processing:
- missingno 0.5.1
- ast.literal_eval
Scientific calculation package:
- Numpy 1.21.5
- Pandas 1.3.4
- Scipy 1.8.0
Machine Laerning tools:
- scikit-learn 1.0.2
- torch 1.10.1+cu113
- torchaudio 0.10.1+cu113
- torchvision 0.11.2+cu113
- tenserflow with keras
Visualization tools:
- Matplotlib 3.4.3
- Seaborn 0.11.2
- Plotly 5.5.0
- Cufflinks 0.17.3
Outline:
I. Data pre-processing
Missing value handling
Movie information extraction
Feature engineering
Addition data modification
II. Data exploration and visualization
Introduce weighted rating
Trend of quantities of different types of films
III. Collabrative Filtering Algorithms
Baseline, SVD, KNN with means methods
Performance comparison
IV. Movie Recommender Implementation
User-rating based recommender
Description based recommender
Keyword based recommender
Hybrid recommender
Deep Learning method
Device Information:
CPU: Intel(R) Xeon(R) Gold 5216R CPU @ 2.10GHZ
GPU: Tesla V100S*2
Estimated running time:
- Data processing part: 5-6 min
- CF part: 2 min
- Deep learning part: 5 min
References: