Regression

Regression is the problem of learning a functional relationship between input features and an output target using training data where the specific functional form learned depends on the choice of model. The parameters of the function are learned using data where the target values are known, so that the machine can make predictions about data where the target is unknown. The goal of a regression model is to learn to predict an output based on an input set of features. The following figure depicts a sample supervised learning model where the training data (depicted in blue) is used to learn a function (depicted in pink) which can then be used in the future to make predictions.

Creating regression models is easy with GraphLab Create! The regression toolkit implements the following models:

• Linear regression
• Boosted Decision Trees

These algorithms differ in how they make predictions, but conform to the same API. With all models, call create() to create a model, predict() to make predictions on the returned model, and evaluate() to measure performance of the predictions. All models can incorporate:

• Numeric features
• Categorical variables
• Sparse features (i.e feature sets that have a large set of features, of which only a small subset of values are non-zero)
• Dense features (i.e feature sets with a large number of numeric features)
• Text data
• Images

Model Selector

It isn't always clear that we know exactly which model is suitable for a given task. GraphLab Create's model selector automatically picks the right model for you based on statistics collected from the data set.

import graphlab as gl

# Load the data
data =  gl.SFrame('http://s3.amazonaws.com/dato-datasets/regression/yelp-data.csv')

# Make a train-test split
train_data, test_data = data.random_split(0.8)

# Automatically picks the right model based on your data.
model = gl.regression.create(train_data, target='stars',
                                    features = ['user_avg_stars',
                                                'business_avg_stars',
                                                'user_review_count',
                                                'business_review_count'])

# Save predictions to an SArray
predictions = model.predict(test_data)

# Evaluate the model and save the results into a dictionary
results = model.evaluate(test_data)

GraphLab create implementations are built to work with up to billions of examples and up to millions of features. GraphLab Create also provides a wrapper to Vowpal Wabbit, an open source out- of-core learning system that is also known to be fast and scalable.