Page 6 - QuantScan-User Guide

Professional investors only
Regression – Statistical learning
Linear regression predicts returns of the fund with a linear combination of returns of the benchmark.
Decision tree predicts returns of the fund with the Classification And Regression Tree algorithm.
Gradient boosted trees predict returns of the fund by iterative training of decision trees.
Nearest neighbors analyze the nearest neighbors in the feature space to predict returns of the fund.
Neural network is composed of layers of neurons which it processes to predict returns of the fund.
Random forest uses an ensemble of decision trees to model the probability of classifiers.
Y (hat) is the predicted return of the fund versus observed returns of the benchmark (respectively Min, Mean and Max return of the benchmark).
Mean Squared Error is the mean square of residuals. The lower the MSE, the better the model. While a value close to zero might suggest a perfect fit, one should take into account that only four significant digits are shown in the table (hence, the mentioning of SD as well).
Mean Absolute Deviation is the mean absolute value of residuals. The lower the MAD, the better the model.
Standard Deviation is the squared root of MSE. The lower the SD, the better the model.
Perplexity is a measure of error of the statistical learning model. The lower the perplexity, the better the predictions made by the model.
Regression – Fund prediction
The blue dots on the graphs show the pairs of observed fund return and observed benchmark return (the orange line indicates the expected fund return as a function of the benchmark return).
6/8