Machine Learning is "a set of algorithms that can make and improve predictions or behaviors based on data."
One of the earliest implementations of Machine Learning (ML) was in image processing. For example, being able to identify pictures that have a dog in them.
Machine Learning is further subdivided into Supervised and Unsupervised models. Supervised ML models are based on a dataset where the outcome is known. Datagration's ML tool can build supervised machine learning models, and use those models to make predictions.
Statistical Modeling (MVR) vs. ML
Multi-Variable Regression (MVR) and ML can produce very similar results, even though their foundations are very different.
MVR uses statistical methods to relate data
ML uses computer algorithms to relate data
Both techniques relate data to an outcome, but use different approaches.
Inference and Prediction
Inference: what can we learn from the data; for example, the data show that a well produces more if it has a longer lateral length.
Prediction: uses the model to make a prediction using a certain set of inputs; for example, using a simple linear regression, y = mx + b, to predict y for a given x.
Regression analysis is simple to build and visualize when there is one x variable (a feature) to relate to y. However, this becomes increasingly complex when you relate y to multiple x features. At best, we can graphically visualize a relationship between y and two x features, as shown in the picture below.
However, if we include three or more features in our model, it becomes much more difficult to visualize all the relationships. Using the Datagration ML tool, users can easily build models that have many features, evaluate the quality of that model, and use that model to make predictions.
Note: The Datagration ML modeling tool has no limitations on the number of features (x's) that can be included in the model. The Datagration MVR tool currently has a limit of 10 features in a model.
MVR and ML Comparison of Capabilities
As mentioned earlier, MVR and ML have the same goal (creating a predictive model), but use different approaches. The table below compares MVR and ML capabilities, modeling techniques and pre-processing options.