LSTM model that forecasts next-second network throughput from the previous 30 seconds of traffic.
- Role
- Data Scientist
- Section
- Data Projects
- Built with
- Python / LSTM / Jupyter
How it works
Packet captures from a Kaggle network traffic dataset are summed into one-second throughput and log-transformed to tame spikes. A 30-second sliding window feeds a stacked bidirectional LSTM (64 then 32 units). The model trains on mean absolute error so traffic bursts do not dominate the fit.