Project Description
Using Keras model from Tensorflow library, the model is trained on historical high price movement of Apple stock. The train and test datasets are created as pairs of current high price and the high prices of the past 30 days. This way, the model is presented with last months prices and is supposed to estimate future high prices based on the firing of simulated neurons inside the neural net.
Even despite the fact that there are numerous other factors influencing the price and also that the preceding price has very little influence on future price, it was still very interesting to create a model that could estimate anything at all. I believe that the model could be somewhat accurate in certain scenarios. For example, if the scope of the data is changed to evaluate prices within a day instead of a month, there could be some recognizable patterns emerging.
Takeaway
Since it has already been more than 20 days from creation of this model it is safe to say that the model largely exadurated the market potential. As such, it is not suitable for production use and probably needs much more data.
