BSR AI Project for Waste Collection Optimization

AI-Supported Analysis of Municipal Cleaning Data for Route Planning Optimization

Project Description

This project analyzes and optimizes the waste collection processes of Berlin's municipal waste management company (BSR) using AI methods. Based on BSR datasets from the past four years, predictions for the upcoming year were developed. To improve prediction accuracy, various publicly available datasets were incorporated into the analysis, including weather data, school holidays, public holidays, and election surveys. The detailed results and analyses are summarized in the project documentation.

Project Structure

• Daten/: Contains raw data and processed datasets
• notebooks_training/: Jupyter notebooks for AI model training
• Notebooks_vorverarbeitung/: Jupyter notebooks for data preprocessing
• Dokumentation.pdf: Comprehensive project documentation
• Leitfragen_Bericht_PPT.pdf: Presentation materials and guiding questions
• Notizen.md: Additional project notes and data descriptions

Data Structure

The project works with the following main datasets:

Base Data

• Month, CW, Year, Date: Temporal recording of deliveries
• Yard: Assignment to different BSR yards (VMF, VMG, VMN, VMM, VMWSF, VMWSM, VMWSN)
• Shift: Working time recording
• Tour: Tour numbers with rotating routes (2-week cycle)
• Tonnage: Waste amount in tons
• Waste Type: Categorization (BIO, HM, SPM)

Additional Datasets

• Weather data
• Inflation data
• Collection failure statistics
• Public holiday data
• Election data
• Tourism data

Installation and Usage

1. Clone repository:

git clone https://github.com/s-matthies/ai-waste-analytics.git
cd bsr_ki

2. Open notebooks in your preferred Jupyter environment
3. Process data according to documentation

Project Context

This project was conducted as a four-day student project by a team of four. It aims to improve waste collection efficiency through data-driven decisions.

Authors and Contributors

• Team of 4 HTW Berlin students