Skip to content

Latest commit

 

History

History
70 lines (70 loc) · 3.45 KB

README.md

File metadata and controls

70 lines (70 loc) · 3.45 KB
Raw

Data analysis and prediction on book data from TIKI

Project structure

D:.
│   .gitignore        # Git ignore file
│   changelog.txt     # Changelog
│   README.md         # This file
│   Report.pdf        # Report
│
├───.ipynb_checkpoints           # Checkpoint folder
│       preprocessing-checkpoint.ipynb # Checkpoint file
│
├───column                      # Column folder: explain the meaning of each column
│       Columns.csv             # Column file
│       Columns.xlsx            # Column file 
│
├───evaluating                  # Evaluating folder: contains the evaluating results
│       all_scores.csv
│       all_scores_2.csv
│       result.csv
│       scores.csv
│       scores_2.csv
│
├───features                    # Features folder: contains the features of the model (after preprocessing and extracting) 
│       processed.csv           # Processed file
│
├───full_data                   # Full data folder: contains the full data after crawling
│       data.csv                # Full data file
│       part_1.csv
│       part_2.csv
│       part_3.csv
│       part_4.csv
│       part_5.csv
│
├───id_data                     # ID data folder: contains the ID data 
│       books_id.csv
│       categories_id.csv
│       id_df.json
│
└───notebooks                   # Notebooks folder: contains the notebooks
        crawl.ipynb             # Crawl data from TIKI
        models.ipynb            # Models
        preprocessing.ipynb     # Preprocessing
        question.ipynb          # Answer the questions

I. Data crawling

  • Crawl data from TIKI via API using several libraries: json, requests, pandas, time, tqdm, ... then save the data to csv files.
    1. Crawl id data: books_id.csv, categories_id.csv, id_df.json
    2. Crawl each part of data: part_1.csv, part_2.csv, part_3.csv, part_4.csv, part_5.csv
    3. Merge all parts of data to data.csv

II. Data preprocessing

  • Preprocess the data: remove duplicates, remove outliers, remove unnecessary columns, rows, use regex to extract features, ...
  • Save the processed data to csv file.

III. Data analysis and answer the questions:

  • Why are there duplicate names of books?
  • How does the low average rating affect?
  • Are book covers and editions of books a significant factor that customers prioritize when hunting for books?
  • How are books that receive a lot of customer attention and reviews typically promoted or offered with discounts?
  • Do the number of pages and the book cover have an impact on the pricing of books?
  • What are the current trends and conditions of books in today's market? => Answer the questions by using data analysis and visualization.

IV. Data modeling

  • Use the processed data to build models.
  • Perform categorical and numerical analysis.
  • Identify the features that have the most impact on the discount percentage of books.
  • Observe the distribution of the features to determine the ouliers and data centrality.
  • Split the data into training and testing sets.
  • Define a pipeline to preprocess numerical and categorical data.
  • Final, use the processed data to build some models (Linear Regression, Random Forest, Gradient Boosting, TranformTargetRegressor) and choose the best model and best hyperparameters by using RandomizedSearchCV.
  • Evaluate the models by using MAE, MSE, ... and save the results to csv files.