Machine Learning Model

‍What is a Machine Learning Model?

A model in machine learning is an object trained to identify specific types of patterns. The ML model is developed by running an ML algorithm on training data. It includes numbers, rules, and algorithm-centric data structures used to make predictions.

An ML model is trained over a set of data to learn from those data and make predictions. Once an ML model is trained, it can be applied to infer over data that has not been seen before and drive predictions about new data introduced.

The best analogy is to think of an ML model as the mathematical representation of a real-world process or a program.

For example, an ML model for credit risk scoring can score applications using previous scores and potential default risk. The natural language processing model is applied to translate words and sentences.  

Taxonomy of Machine Learning Models

Components of an ML model

Mathematical function

These are mathematical equations derived from the data. For example, Y=mX+b is a function used for a simple linear regression model function.

Output variable

It is also termed as a response or dependent variable. ML models predict the value for this variable.

Input variables/features

These variables are also called independent or predictor variables. These variables offer input data to feed into the ML model and predict the output variable.

Parameters/coefficients

These represent coefficients of mathematical equations used in the ML model. ML algorithm learns these based on the historical data and the loss or cost function. For example, weights and biases of a neural network or the coefficients of a logistic regression model. ‍

Hyperparameters

Hyperparameters are the ML model’s initial settings configured before training machine learning models. These are used with the cost or loss function to estimate coefficients discussed above. For example, the number of hidden layers in a NN, train-test split ratio, and choice of activation function in a neural network.

Loss/cost or objective function

The loss function is used to check how accurate the model is but is mainly used to create gradients that correct the parameters so that subsequent performance is better. Optimized loss function helps select the most appropriate parameters and hyperparameters.  

Steps Involved In ML Model Building 

ML projects involve the following steps: 

• Data collection
• Data preparation and processing
• Feature engineering
• Feature selection/extraction
• Model building
• Model validation
• Model selection
• Model deployment
• Model monitoring and retraining based on performance

Further Reading

What is a machine learning model?

Difference Between Algorithm and Model in Machine Learning

Parameters and Hyperparameters in Machine Learning and Deep Learning