I have trained the time series prediction model using old data, and as time goes by, I get more data and want to update my model with it. But it seems like my model perform worse as I update the model.
Here is what I did. First I trained with the existing data.
model = build() model.compile(optimizer='adam', loss='mse') model.fit(X_old, y_old, validation_data = (X_old_val, y_old_val), epochs=100, batch_size=32, verbose=0, callback = earlystopping) model.save_weight(path)
I used this model for a while, and then I gathered new data and update the model.
model = build() model.load_weight(path) model.compile(optimizer=adam, loss='mse') # update the model model.fit(X_new, y_new, epochs=1, batch_size=32, verbose=0)
When I update the model, I do not have validation/test data as I wanted to use the most fresh data for time series prediction. To avoid overfitting I pick small epochs varying from 1-3.
But somehow my model performs worse than before. Does anyone have an idea why this is happening? How I should remedy this?