Water Mains Breaks Predictions

Final project for SI 670, Applied Machine Learning, at the University of Michigan Group members: Coulton Theuer, Bella Karduck, Haley Johnson

Objective

This project tries to predict if a water main will break in the next 3 years in Ann Arbor, Michigan.

These predictions can then potentially be used by the city to mitigate the repair costs, flooding, and property damage associated with water main breaks and help the city better allocate scarce resources for infrastructure improvements.

Data

Data was provided to us by the City of Ann Arbor and can be found in data/raw

Key datasets include:

• All_Watermains_Attributes.csv: information about all water mains in the city's distribution system
• Watermain_Attributes_Soil_Ph.csv: soil information and work order IDs for water mains that have broken in the past
• Watermain_Breaks_Reporting.csv: breaks reported to the City of Ann Arbor between 2013-2023

Directory Structure

---

The directory structure of your new project looks like this:

├── README.md          
├── data
│   ├── transformed    <- Transformed data, include training and testing sets 
│   └── raw            <- Raw data from the City of Ann Arbor
│
├── models             <- Model traning & development
│   └── 01_baseline    <- Baseline model for comparison 
│   └── 02_models      <- Scratch notebook for developing models
│   └── 02_models_cleaned    <- Cleaned up model code, implement final random forest predictor
│   └── 02_svm         <- SVM model, attempting to replicate results of Syrcause paper
│   └── top_152.csv    <- 15 pipes that are most likely to break in the next 3 years
│
├── notebooks          <- Data mainpulation & exploration 
│   └── 01_EDA         <- Combining datasets & exploratory analysis
│   └── 02_train_test_val_split   <- Split data based on time cutoffs
│   └── 03_svm_model   <- Ignore, saved to notebooks by accident
│
├── syrcause_public   <- Cloned repository from the Syrcause paper we used as inspiration for our project
│                        We did not use this code in our project, it was mostly for our refernece
│
├── src                <- Source code for use in this project.
│   ├── __init__.py    <- Makes src a Python module
│   │
│   ├── utils          <- Helper functions to process data for train/test files
│   
└──