01 / THE PROBLEM
Start with a clear question.
I wanted a practical ML project with a place, a dataset, and an output that could be explored. ForecastML uses historical Kolkata weather data from 2017–2022 for a seven-day forecasting experiment.
02 / DESIGN DECISIONS
Work with historical weather observations rather than inventing sample values.
Use RandomForestRegressor as an approachable baseline for a regression task.
Use Plotly to make the output easier to inspect than raw terminal numbers.
Treat the result as an experiment; meaningful forecasting needs proper validation and comparison with baselines.