Baltimore Crime data

Baltimore

Baltimore crime forecasting

I have recently started learning time series and a fascinating technique that I have learnt was Holt-Winters method. It is an exponential time series that accounts for residuals in decreasing priority. More info here.

It is good for short-term forecasting. As such, I made a 1 month forecast and noticed a downward trend. However, there are some improvements to be made:

• Outlier detection (eliminating outliers to make less variant model)
• Comparative analysis (with ARIMA, VectorAutoregression, LSTM, XGBOOST etc.)
• In-sample/Out-sample prediction
• Diagnostic analysis (to tune the model better)

Here is my code