Skip to content

Latest commit

 

History

History
80 lines (51 loc) · 3.07 KB

README.md

File metadata and controls

80 lines (51 loc) · 3.07 KB

Class Profile 1A Data Analysis (with ML)

The class profile is an analysis done on students in an engineering cohort (in this case, Systems Design Engineering 2025) to display information about the class to the general public.

This repo contains the analysis for sections that used machine learning algorithms to find insights in the data

Note: This is not the full repo; this repo only contains data analysis done by me using ML. To see other visualizations created by me and the rest of the team, click here

Motivation

  • Past engineering cohorts at the university of Waterloo had already done similar projects (analyzing class data)
  • To add another layer of interest to the analysis, me and my team decided that we would do a section of analysis using ML algorithms to extract insights from data
  • I wanted to put my ML skills to the test and find new things in the data... which I was able to do!

Tech/Framework


Analysis Done Using
  • Python
  • Sklearn (TSNE, KMeans, PCA)
  • XGBoost
  • SHAP (for creating feature importance graphs)
  • Yellowbrick (visualizing optimal clusters for Kmeans)
  • Optuna (hyperparameter tuning for XGBoost)
  • Seaborn (graphs)
  • Numpy
  • Pandas (data manipulation)

Highlights

  • Cleaning survey data using pandas, sklearn, and custom cleaning functions
  • Hyperparameter tuning of an XGBoost model using Optuna
  • Extracting insights from a trained model using SHAP
  • Visualizing higher dimensional structures in data related to courses using TSNE
  • Finding optimal numbers of cluster for kmeans through elbow method and silhoutte scores
  • Clustering 4D course data to find data points in similar clusters

Notebooks

  1. ML/xgb_feature_finder.ipynb for my work on finding factors that predict academic success (XGBoost + Shap)

  2. ML/course_by_person.ipynb for my work on finding similar courses by clustering (TSNE + Kmeans)

Files

  1. SYDE_2025_ML_Report_Nicolas.pdf for report containing results with insights explained

Screenshots

Refer to graphs folder if images do not load

Note: For analysis of the graphs, refer to SYDE_2025_ML_Report_Nicolas.pdf in the repo

  • Feature Importance Graphs

SHAP 1

SHAP 2

  • TSNE and Kmeans course Clustering

TSNE by Course

TSNE by Cluster

Author

  • This data analysis was done by me - an aspiring data scientist who is always curious and looking to find insights in data
  • I would also like to thank Sean for introducing me to some of these concepts, as well as Nirmal and Alan for working with us on the data analysis for the rest of the project