Tesla Stock Analysis

A reproducible educational research artifact for market signals, forecasting experiments, and sentiment analysis.

This repository explores historical Tesla-market behavior through technical indicators, machine-learning forecasting, and sentiment signals. It is an educational research project intended to make the workflow easy to inspect and reproduce in Jupyter or Google Colab.

Important: This is not financial advice. Historical performance and model predictions do not guarantee future results.

What this project demonstrates

Research layer	Included work
Market data	Historical Tesla price analysis for 2015–2020
Feature engineering	Technical indicators and social/sentiment features
Forecasting	XGBoost and LSTM experiments
Interpretation	Exploratory visualizations and sentiment analysis
Reproducibility	Pinned dependencies, notebook validation, and documented data boundaries

Start here

Goal	File
Review the main forecasting experiment	Hossein_Dehghan_Tesla_Stock_Prediction_XGBoost_LSTM.ipynb
Explore historical behavior	tesla_stock_analysis_Hossein_Dehghan_.ipynb
Inspect the sentiment workflow	colab_sentiment_code.ipynb
Recreate the sentiment plot	recreate_sentiment_plot.py
Understand data boundaries	docs/DATA_PROVENANCE.md
See the broader engineering portfolio	Hossein Dehghan on GitHub

Run locally

git clone https://github.com/hossiendehghan989/Tesla-Stock-Analysis.git
cd Tesla-Stock-Analysis
python -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt
python -m pip install pytest==8.4.2 nbformat==5.10.4
python -m pytest -q
jupyter lab

Open a notebook and run the cells in order. The notebooks may require a locally supplied data snapshot; follow docs/DATA_PROVENANCE.md first.

To recreate the sentiment figure without a machine-specific path:

python recreate_sentiment_plot.py path/to/data.csv --output artifacts/sentiment_over_time.png

How to read the results

The experiments depend on the selected historical period, feature engineering choices, data availability, preprocessing, and model configuration. Forecasting output should be interpreted as an experiment in modeling—not as a trading signal or investment recommendation. A stronger next step is a leakage-audited, walk-forward benchmark with explicit baselines and transaction-cost assumptions.

Research boundaries

The CI workflow validates notebook structure and the portable plotting utility. It does not execute every notebook against a live data source; that boundary keeps CI deterministic and avoids depending on external market-data availability.

Next research directions

• Add a chronological walk-forward evaluation protocol
• Compare against naive and seasonal baselines
• Separate exploratory features from production-eligible features
• Document missing-data handling and train/test boundaries
• Add error analysis by market regime
• Publish a versioned data manifest for each reproducible experiment

Author

Built by Hossein Dehghan, an Industrial Engineering graduate focused on applied AI, data science, energy intelligence, and operational decision support.