Skip to content

Latest commit

 

History

History
83 lines (68 loc) · 4.22 KB

README.md

File metadata and controls

83 lines (68 loc) · 4.22 KB

Predict the relevance of search results from E-Commerce

Kaggle Crowdflower Search Results Relevance data를 이용한 E-Commerce 사용자 검색 시스템 만족도 예측 모델

Description

프로젝트 과정 설명(링크는 Medium글)

Project structure

This project is organized as follows.

.
└── utility/           
    ├── README.md 
    ├── __init__.py
    ├── augment.py                     # Data augmentation function
    ├── eda.py                         # EDA: Easy Data Augmentation Techniques for Boosting Performance on Text Classification Tasks
    ├── processing.py                  # Preprocessing functions
    ├── predict.py                     # predict functions
    └── utility.py                     # metrics, distance stack, plot, etc. functions
└── example/      
    ├── Description.ipynb              # description for this project
    ├── EDA.ipynb                      # Exploratory Data Analysis
    ├── LSA.ipynb                      # The whole process of the project
    ├── preprocessing.ipynb            # Data preprocessing process flow
    └── word2vec.ipynb                 # Implementing and applying word2vec(*Implemented in tensorflow 1 version.)
├── .gitignore       
├── README.md
├── gridsearch.py                      # Parallelized gridsearchCV to find hyperparameters
└── main.py                            # Make submission py

Exploratory Data Analysis

데이터 탐색

  • EDA.ipynb

Preprocessing

데이터 전처리 과정

  • preprocessing.ipynb
  • utility/processing.py

Data Augmentation

데이터 증강

  • utility/augment.py
  • utility/eda.py

model implement

사용자 만족도 판별 모델링 과정

  • LSA.ipynb
  • word2vec.ipynb

utility

모델에 사용된 함수
utility/README.md 참고

Usage

processing.py 로 데이터 전처리후 data Augmentation을 합니다.(data Augmentation의 각 하이퍼파라미터는 논문을 따릅니다) Augmentation main.py로 submission을 생성합니다.

python utility/processing.py --input=./data/train.csv --eda=True
python utility/augment.py --input=./data/eda/train_1.txt --num_aug=8 --alpha=0.05
python utility/augment.py --input=./data/eda/train_2.txt --num_aug=4 --alpha=0.05
python utility/augment.py --input=./data/eda/train_3.txt --num_aug=4 --alpha=0.05
python utility/augment.py --input=./data/eda/train_4.txt --num_aug=0
python utility/processing.py --input=./data/eda_train.csv
python main.py --mode=eda

참고자료

  1. Predicting the Relevance of Search Results for E-Commerce Systems
  2. Using TF-IDF to determine word relevance in document queries
  3. Classifying Positive or Negative Text Using Features Based on Opinion Words and Term Frequency - Inverse Document Frequenc
  4. An introduction to latent semantic analysis
  5. Using Linear Algebra for Intelligent Information Retrieval
  6. EDA: Easy Data Augmentation Techniques for Boosting Performance on Text Classification Tasks
  7. Weighted kappa loss function for multi-class classification of ordinal data in deep learning