Deep learning model is a neural network with multiple hidden layers, more input data, and adding more layers can make a neural network complex to analyze without a gain in performance. Sometimes your model works but in order to make it work effectively change the algorithm of the training network.

You can test several architectures: consider one hidden layer and increase the number of nodes or increase the number of hidden layers (deep learning model) and then change the number of nodes in each layer. To test the algorithm, the correlation and error of training, validation, and test subset should be monitored to avoid over- and under-fitting. There is no general optimum number of hidden layers and nodes for all problems. The architecture of the neural network should be tested for each problem individually.

Have a look at

https://towardsdatascience.com/applied-deep-learning-part-1-artificial-neural-networks-d7834f67a4f6

If your dataset is time series related I recommend to use RNN model (LSTM).