using LSTM to predict the temperature in the future
The dataset I used is from NASA.The data they give include the temperature from 800,000 years ago, even earlier. In my opinion, the temperature from 1880-2015 will be perfect.
Indeed, there are some imperfections in this code. My friend uses tensorboard which is a great tool to visualize the data, including LOSS, prediction data, estimated data, etc.. I also heard that it is easy to conduct but that code should be added in sess. In short term I don't have time to go deep into it.
Tips: If you improve the value of steps, the accuracy will be better. Although the result will be better, the process of analyze the result is miserable. I put all the data in EXCEL to visualize which is stupid.