Logo Lanfrica

kikivanrongen/PredictiveMaintenance_WaterPumps

Domain:

environment and energy

Record type:

project
Creator:
kik
Host:
Machine learning project regarding predictive maintaince for water pumps in Tanzania, based on the data of the Tanzanian Ministry of Water # Goal Building a machine learning model that is able to accurately predict maintenance on water pumps in Tanzania, based on the data of the Tanzanian Ministry of Water # Project overview This project contains the following folders: 1. *data*: containing the data files in csv format 2. *eda*: contains a notebook that is used to perform exploratory data analysis 3. *analyze*: contains files for ETL, feature engineering and training & evaluating the classification model 4. *images*: all images regarding the model evaluation are saved to this folder 5. *results*: folder to store pickle files containing the predictions # Getting started (Docker installation) We launch the Flask application using Docker. Build docker image from root directory with \ `docker build -t .` \ then run the application with \ `docker run -it -d -p 5000:5000 ` # Re-train model If you are interested in reproducing the results, the following commands can be executed from the analyze folder: 1. Extract, transform and load the data with `python etl.py` 2. Feature engineering with `python feature_engineering.py` 3. Train and evaluate the model with `python train_evaluate_model.py` The predictions are stored in the results folder, but are also visible through a REST API, by running \ `python app.py` \ (similar to getting started with Docker)