All projects
Machine learning · 2023
Concrete Strength Prediction
Kaggle Playground Series S3E9 — regression
A Random Forest model that predicts the strength of a concrete mix from its ingredients: the kind of model construction teams use for quality control.
- 13.157
- private score (MAE)
#Overview
I built a model to predict the strength of a compound from its ingredients. Models like this are useful in industries such as construction, where understanding material strength from composition is crucial for quality control.
#Tools
- Pandas for organising and loading data
- scikit-learn for the machine learning algorithms and evaluation
#Method
- Setting up and loading data: loaded the dataset of compound recipes and their resulting strengths.
- Preparing the data: identified the features that could predict strength: the types and amounts of ingredients such as cement, blast furnace slag and fly ash.
- Building the model: chose a Random Forest, which handles complex datasets where many features influence the outcome. The model learned the relationship between ingredients and strength from the labelled data.
- Evaluating the model: compared predictions to true strengths using Mean Absolute Error (MAE), where lower is better.
- Making predictions: predicted strengths for unseen recipes and prepared them for competition submission.
#Results
The Random Forest scored a private MAE of 13.157 on the competition leaderboard, and the same approach can be adapted to many other predictive modelling tasks.