Multi Layer Perceptron neural network with no ML libraries. Uses Genetic Algorithms for optimisation. Written in Python.

• Model has 1 hidden layer with arbitrary neurons.
• Model output layer has 1 neuron.
• Trains with back propagation using gradient descent of the error function.
• Model optimisation is done via a Genetic Evolutionary Algorithm

pip install numpy pandas copy matplotlib warnings deap scipy bitstring

The best results for the river basin's pan evaporation prediction (Fresno, CA):

Validation RMSE: 0.010292

Test RMSE: 0.009849

Denormalised RMSE: 0.158563

Comparison with multiple linear regression model: