Python scripting language?

Based on what is stated in the question, the following conclusions can be drawn:

• Your neural network is likely overfitting. This follows from the fact that you have good results on the training set, but poor results on the test set. One possible solution is to use regularization. Additionally, try using the dropOut method. This technique helps reduce overfitting by randomly dropping connections in the neural network.
• You need to perform cross-validation (cv). This procedure is used to evaluate the model and tune parameters. We are interested in the evaluation. scikit-learn has excellent mechanisms for doing this. One cv option (which is more suitable for you since your sample is small) involves splitting the original dataset into 2 parts.