This project is from the Machine Learning challenge hosted in Driven Data Pump it Up: Data Mining the Water Table where i placed in rank 903 out of 13.000 competitors
What was developed:
Estimator to predict if the water pump is functional, needs repair or non-function: configures a multi label classification model with accuracy score of 82.15%
Exploratory Data Analisys: The data contains a lot of features, so it was essential to visualize every distribution, relation and correlation between variables
Feature Engineering and Importance: using random forest, we can see each feature importance to the estimator. Also, feature engineering was extremely import to process non-numerical features, along with the creation of new features using the spatial ones that the data contains (latitude, longitude)
Several models where used, including Random Forests, Neural Network, and CatBoost, with the last being the best fit and accuracy score
After the modeling i have build a deploy for identifying in which category the water pump fits. The deploy is a map visualization of Tanzania, using the Plotly Python library.
For building and maintaining the app online, i used Streamlit and Streamlit Cloud.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent