Pollution

This roject pretends to forecast 24 hours Pollution for London City using NN. The selected pollutant was NO2 and the selection criteria can be found in the project LdnAirAnalysis

GRU - LSTM One step univariate model to forecast NO2

The present project uses a dataset previously cleaned and analyzed in the project LdnAirAnalysis. The GRU and LSTM NN will predict pollution for the next hour and indicate an alarm according to the European Environmental Agency EEA.

Data Preparation

The imput data contains the values of the previous 24 hours T=24 of NO2 concentrations. The choice of T=24 was random but should be selected through experimentation. Since the aim is to forecast the next hour, is necesary to define HORIZON=1 to specify the forecasting horizon of 1 (t+1)

Plots

Overview of the dataset

Training Vs Validation Vs Test

LSTM Evaluation Table and Graph