Seoul_Bike_Sharing_Demand_Prediction

Topic: Predicting demand for rented bicycles around Ewha, Seodaemun-gu and Mapo-gu

Environment Setting

• Google Colab
• model name : AMD EPYC 7B12
• number of CPU cores : 2

Specification of memory(RAM)

• MemTotal : 13297228 kB
• MemFree : 7424924 kB
• MemAvailable : 11290352 kB

Dataset

• Original Dataset:

  • “공공자전거 대여소 정보(22.06월 기준).csv” from 서울 열린데이터 광장(Seoul Open Data Plaza)(https://data.seoul.go.kr/)
  • “서울특별시 공공자전거 이용정보(시간대별)_21.01.csv”~“서울특별시 공공자전거 이용정보(시간대별)_21.12.csv” from 서울 열린데이터 광장(Seoul Open Data Plaza)(https://data.seoul.go.kr/)
  • “SURFACE_ASOS_108_HR_2021_2021_2022.csv”(서울특별시 2021년도 종관기상관측자료) from 기상자료개방 포털(KMA) (https://data.kma.go.kr/data/grnd/selectAsosRltmList.do?pgmNo=36)

• Original Dataset is too big, so we preprocessed data and reduced the size of dataset

  • we dropped correlated features of weather data. and merge it to bike sharing data. you can see that process at '/original data/dataset.ipynb'
  • Our Dataset: seodaemunAndmapo.csv (72,463KB) (Total number of this dataset: 972799)

Models

• RandomForest and Gradient Boosting (In this project, the result of gradient boosting was better than random forest)
• Hyperparameter Search: GridSearch
• After Hyperparameter tuning, R2 Score of Gradient Boosting was 0.9380330070944691

To Run this project

1. Download notebook file. (seoul_bike_sharing_damand_prediction.ipynb)
2. Download dataset file(seodaemunAndmapo.zip), unzip this file, and get seodaemunAndmapo.csv file.
3. Open notebook file in Colab and upload csv file to Colab or your Google Drive.
4. Run cells

• The features of our datset is korean. so, you need to install fonts following the instruction at the top of the notebooke file.

# For Hangeul Font Issue
!sudo apt-get install -y fonts-nanum
!sudo fc-cache -fv
!rm ~/.cache/matplotlib -rf

run this cell and restart runtime

• If you downloaded project files on your drive, then Mount your Google Drive.

# Mounting Google Drive 
from google.colab import drive
drive.mount('/content/drive')

• set the data path(location of seodaemunAndmapo.csv) to DATA_PATH

*. the path should be absolute path

# set data path
import os
DATA_PATH = '/content/drive/MyDrive/seodaemunAndmapo.csv'

• run cells to load dataset, data preprocessing, train model, evaluation, interpretation

References

• kaggle (https://www.kaggle.com/code/kwonyoung234/for-beginner/notebook)
• scikit-learn (https://scikit-learn.org/)
• matplotlib (https://matplotlib.org/)
• heumsi's github (https://github.com/heumsi/Seoul-Public-bicycles-EDA)