Predicting The Next Day's Trend with Machine Learning

I have to make a short detour. Remember from last post that my next assignment is to create the trade back testing application? Well, I have set it aside for a while to give way for this new function. 

I have added a function to my python desktop application which is  also part of my original plan. This new function allows me to predict the next day's trend(whether the ticker will go down(Sell) or  up(Buy)). This is on its beta version. I have tested it with 30 days using historical data and along the fine tuned it on order for me to yield acceptable result. So far I got 85% accuracy. here is the screenshot of the new interface:

The code is not mine, I have based it from NeuralNine. Here is the video tutorial:

NeuralNine used Sequential model having LSTM(Long Short Term Memory) layers. He intentionally predicted the exact closing price of the following day which is quite impossible to do, he achieved 52% accuracy which is very low, but when I checked the chart, it seems that he can predict the trend more accurately than predicting the exact closing price that is why I decided to give it a try.

With this new function, my confidence level will go up by 5% because of the 85% accuracy, once I have achieved 98% the confidence level will go up by another 5%.

I will be testing this new function everyday and at the same time continue adding new functions to my desktop app. My plan is to redesign the layout (future plan) and probably the next assignment will be the technical analysis prediction. I want to work on this first(Technical Analysis Prediction function) because I feel that it is very useful(more useful than the trade back testing application).