Stack Overflow 2023 Survey — Power BI & Streamlit Visualization & Prediction Model

(For Streamlit project please seek streamlit folder in this repo)

This repository contains a Streamlit app also a Power BI project and an accompanying prediction model built from the Stack Overflow Developer Survey 2024 data. The Streamlit app (see the streamlit folder) provides interactive visualizations; the repo also includes the Power BI report file (ProjeST.pbix), an exported Power BI PDF (ProjeST.pdf), a project report PDF, and a Jupyter notebook (PredictionModel.ipynb) with the prediction model and experiments.

Files of interest

• streamlit folder - Includes all files and app.py related to streamlit project.
• ProjeST.pbix — Power BI report (Power BI Desktop file).
• ProjeST.pdf — Exported PDF of the Power BI report (embedded below).
• PredictionModel.ipynb — Jupyter notebook with the prediction model and experiments.

How to view

• To view Streamlit Vizulation please view README.md that inside streamlit folder.

1. PDF: contains pages of PowerBI.
2. Power BI: open ProjeST.pbix with Power BI Desktop (Windows).

---

Project report (Power BI)

Figure: PowerBI dashboard overview page 1 of 4

---

To view all pages in PowerBI, please seek ProjeST.pdf.

---

Prediction model

The PredictionModel.ipynb notebook contains the data-preparation steps, model training, and evaluation used to predict outcomes from the Stack Overflow survey data. Open it with Jupyter or VS Code's notebook support.

Notebook summary

The PredictionModel.ipynb notebook contains the model pipeline and experiments used to predict annual developer compensation from the Stack Overflow survey. Main sections:

• Lib Imports — Import standard data science libraries and configure warnings.
• Data Fetch — (Optional) download dataset with kagglehub and load the CSV.
• Basic Data Clean — Selects relevant columns and filters salary outliers.
• Feature Preprocessing | Data Augmentation — Convert and simplify features (experience, education, dev type, country, employment) and keep top countries.
• Pipe Line — Build preprocessing pipelines (imputation, scaling, one-hot encoding) and a GradientBoostingRegressor; includes evaluation and plotting helpers plus a main runner.
• Model Run — Calls main(df) to execute the full training/evaluation flow.

Quick start (run the notebook)

1. You can easily run model on Google Colab.