Weather-Forecasting

Using a scratch LSTM, being able to predict the temperature

Math_LSTM file

In this pdf you will find my derivation for the backpropagation equations, equations needed in order to implement a scratch LSTM. If you want to follow along it is necessary some knowledge of Linear Algebra and Calculus.

LSTMClass

Using the forward propagation and backpropagation equations derived in the Math_LSTM file, creating a LSTM recurrent neural network using only the numpy library.

Code folder

In this folder there are two jupyter files:

1. Simple application. In this notebook you will find an application of the LSTM, a very simple one: predicting the next value of a sinusoidal function.

2. Clime forecasting. Although the application is the same as in the previous file, predicting the next value of a given sequence, in this notebook I have loaded and preprocess the data in order to make predictions on the temperature. It also includes a comparison with the TensorFlow LSTM model in order to see how well the scratch implementation works.

Final Results on the Jena dataset

As we can see in the image, the scratch implementation is able to predict the temperature almost as good as the tensorflow model.