The table below shows the estimated model for predicting the Sepal Length using a generalized linear model with Elastic-net Regularization using Sparks native machine learning library.

Parameters

Select a set of input variables below to predict the sepal lenght of flowers

This application predicts the output of the designated response variable using a deep neural network. The model can have up to four hidden layers. The model also runs a correlation analysis, removing non-linear variables from the neural network input parameter list. In short, this acts as a way to speed up calculation. Careful though, a high threshold will throw an error if set too high. This also enables the network to define its own inputs and layers. The user can adjust the number of available hidden layers, correlation threshold, learning replications, and learning rate.

Two model are actually ran. One model is produced by the neural network and the second is a linear regression model. This is to help understand the performance from both approaches. The plots below shows how they compare to each other in terms of accuracy. The black line on the bottom plot assumes the ideal case and that both models produced the same values. The blue line is the deviation from that. The bias is determined from that difference. The mean squared error is displayed on the top plot from both models.