OBD-II-Eco-driving-assistant

This repository contains codes and datasets in correspondence to my bachelor thesis for computer science at Vrije Universiteit Amsterdam.

Datasets

• exp1_14drivers_14cars_dailyRoute.csv -> original data set for the experiment, acquired from the public dataset published by Cephas Barreto.

• exp2_19drivers_1car_1route.csv -> same as the previous one.

• train_shuffled_16features.csv -> the train set with 16 features after initial data engineering.

• train_shuffled_7features.csv -> the train set with 7 features after selecting important features.

• train_shuffled_43features.csv -> the train set with 43 features after feature crosses.

• test_shuffled_16features.csv -> the test set with 16 features after initial data engineering.

• test_shuffled_7features.csv -> the test set with 7 features after selecting important features.

• test_shuffled_43features.csv -> the test set with 43 features after feature crosses.

• total_16features.csv -> the total set with 16 features before spliting corresponding train and test set.

• total_7features.csv -> the total set with 7 features before spliting corresponding train and test set.

• total_43features.csv -> the total set with 43 features before spliting corresponding train and test set.

• case_study_19drivers.csv -> dataset used for the case study mentioned in Chapter 4 of the thesis.

Notebooks

• data_preparation_xfeatures.ipynb -> data wrangling and engineering to produce the corresponding dataset.

• data_preparation_case_study.ipynb -> data wrangling and engineering to produce the dataset used in the case study .

• modelling_xfeatures.ipynb -> build regression models according to datasets with different number of features.

• modelling_case_study.ipynb -> case study for 19 drivers as mentioned in Chapter 4 of the thesis.

• automl_1h.ipynb -> 1h training using automl, the dataset and regressor used can be modified accordingly

Running environment

All codes are written in python and jupyter notebook. Jupyter notebook can be installed following the official installation guide.

After successfully deployed the jupyter notebook, all notebooks can be compiled directly on python3 for a version higher than 3.9. When compiling, make sure that all documents are in the same directory. There is no requirement for the operating system, with the exception of automl_1h.ipynb.

Automl requires python3 and jupyter notebook to be deployed on a Linux operating system. In order to run automl in other operating systems, possible solutions are WSL for windows, virtual machine, docker image, etc.