Project - Machine Learning Analysis of Train stations

• Pierre Jézégou
• Julie Oppedal

Project

Understanding how many passengers use train stations and what features they have is important for improving transportation services. We’re using machine learning to analyze train station data and predict passenger traffic.

To start, we’re gathering data from reliable sources like the SNCF and INSEE websites to create a detailed dataset with lots of information about cities and train stations. Next, we’re cleaning up the data to fix errors and make sure it’s good for analysis. Then, we’ll use different machine learning models to try and figure out the best way to predict passenger traffic and understand what factors make some stations busier than others.

Overall, this project focuses on using advanced computer techniques to optimize transportation services by accurately predicting passenger traffic and understanding station dynamics.

Documents

• project_proposal: first deliverable
• main: final report

Project organization

.
├── README.md
├── requirements.txt
│
├── documentation
│   ├── BUILD_DATASET.md
│   ├── NOTICE.md
│   ├── TODO.md
│   └── project_proposal.tex
│
├── data
│   ├── attraction_area_per_city.csv
│   ├── cities_statistics.csv
│   ├── frequentation-gares-raw.csv
│   ├── gares-de-voyageurs.csv
│   ├── gares-equipees-du-wifi-raw.csv
│   └── gares-pianos.csv
├── build-dataset
│   ├── build_dataframe.py
│   └── consolidate_cities.py
│
├── Preprocessing.ipynb
├── distance_calculation.py
├── dataRead_processed.pkl.bz2
├── dataset_preprocessed.csv
│
└── models
   ├── helpers.py
   ├── knn.ipynb
   ├── linear-models.ipynb
   ├── metrics.csv
   ├── randomforest.ipynb
   ├── sampling.ipynb
   └── svm.ipynb 

• documentation: contains all the documentation files
  • BUILD_DATASET.md: explains how to build the dataset manually (without the CI)
  • NOTICE.md: contains the notice for the project, to reproduce the results
• Build dataset
  • data: contains all the data files
  • build-dataset: contains the scripts to build the dataset
• Preprocessing notebook and algorithms are located at the root of the project
• models: contains the notebooks for the models

Usage

• Install working environment

  • Install virtual environment (venv): python3 -m venv .venv && source .venv/bin/activate
  • Install dependencies & libraries : pip install -r requirements.txt

• Train the model

• Use the model

Features

Feature	Description	Data Type	Number of Unique Values
city_name	Name of the city	string	String
Trigramme	Trigram of the city	string	String
drg_segment	DRG segment	string	Categorical
geographical_position	Geographical position	string	String
city_code	City code	string	String
uic	UIC code	string	String
postal_code	Postal code	string	String
total_passengers_2022	Total passengers in 2022	integer	Integer
total_passengers_and_non_passengers_2022	Total passengers and non-passengers in 2022	integer	Integer
Total Voyageurs 2021	Total passengers in 2021	integer	Integer
total_passengers_2015	Total passengers in 2015	integer	Integer
total_passengers_and_non_passengers_2015	Total passengers and non-passengers in 2015	integer	Integer
piano	Piano	string	Categorical
power_station	Power station	string	Categorical
baby_foot	Baby foot	string	Categorical
distr_histoires_courtes	Distribution of short stories	string	Categorical
total	Total	string	Categorical
wifi_service	WiFi service	string	Categorical
Unnamed: 0	Unnamed column	string	Categorical
city_label	City label	string	Categorical
city_attraction_area	City attraction area	string	Categorical
city_attraction_label	City attraction label	string	Categorical
department	Department	string	Categorical
region	Region	string	Categorical
sum_municipal_population_2021	Sum of municipal population in 2021	integer	Integer
non_scholarized_15_years_old_or_more_2020	Number of non-scholarized individuals aged 15 or more in 2020	integer	Integer
main_residences_2020	Number of main residences in 2020	integer	Integer
housing_2020	Housing in 2020	integer	Integer
jobs_at_workplace_2020	Number of jobs at workplace in 2020	integer	Integer
active_employers_2021	Number of active employers in 2021	integer	Integer
hotels_2024_sum	Sum of hotels in 2024	integer	Integer
hotel_rooms_2024_sum	Sum of hotel rooms in 2024	integer	Integer
camping_sites_2024_sum	Sum of camping sites in 2024	integer	Integer
camping_sites_2024_sum.1	Sum of camping sites in 2024	integer	Integer
other_tourist_accommodations_2024_sum	Sum of other tourist accommodations in 2024	integer	Integer