# Sonja Model (Sets 01 13) Zip

using the same data from above you can also train a model with a very small data set. this data set consists of just 13 examples (one per month) and contains no random splits. the target is a univariate time series with a length of just 12 time steps

the input is a multivariate time series with a length of 12 time steps and 3 dimensions

the initial timesteps is set to 0 and does not change

prepare the data this model is trained using the very same data. model configuration model structure the model consists of one linear layer of size 1x12x3. it has a linear activation function. model hyperparameters model training is using a batch size of 32 and a batch momentum of 0.95. the learning rate is set to 0.05. output interpretation model training produces an output of 1x1x12x3 indicating the prediction for each example. model results the model has a mean squared error of 0.0448. sveta model (sets 01 13) zip training a model with an untuned learning rate and a very small data set can result in a poorly performing model. model configuration the model consists of two linear layers of size 1x12x3. model comparison you can use the models you have trained to predict the returns of the next month.

the code consists of six classes. the first is the main sonja class which provides a constructor and main method which takes as arguments: s01_model s01_model_attributes and returns: s01_model_attributes which includes a bunch of useful data about the model (like the number of attributes, the number of classes, the percentage of data set in the training and test sets, the percentage of the training set data which is contained in the training set, the training and test error rates, the confusion matrix and other useful bits and pieces). the second class is the defaultattributes class which is used to provide a data set for the model to use. the third class is the class to implement the svd algorithm. the fourth class is the class to implement the decision tree algorithm. the fifth class is the class to implement the attribute selection algorithm. the final class is the class to output the results from the training and testing data.

## Sonja Model (Sets 01 13) zip