Machine Learning System

Introduction To Machine Learning System

The machine learning system is an overall picture of the Machine learning use case for real-world problem scenarios. Machine learning is not only about building the prediction model but using them in real-life problems. To make use of the machine learning model it shall be deployed and can be accessed via client requesting via APIs. Model is trained on already existing historical data, but for inference, there could be several ways, it can be fed with the whole set of data or small chunks fed via API calls and show it to the user at the front end.

What Is Machine Learning?

“Machine learning (ML) is a field of Science & Technology, which makes use of mathematics & statistics for algorithms to understand the data and learn about features by assigning a value (Coefficient) to each independent variable without hard coding any explicit logic. Machine Learning algorithms find the available patterns in data and based on that predicts the outcome.”

Machine Learning System

In real-world machine learning needs to be incorporated with a large set of components of the application. ML code is just one of the components as can be seen from the below figure, which just contributes to more or less 5% of the overall system.

Machine Learning System, Source: developers.google.com

A large and considerable amount of time is spent in collecting the input data, performing sanity checks on the data, and extracting important and explainable features from it. A serving infrastructure must be built to place the model predictions to be used in the real world.

Well building the whole system is not the job of a data scientist alone, it’s an end-to-end process and many developers are required to make it work.

How To Build A Machine Learning System?

The job of a Machine learning engineer or data scientist is to come up with a scalable solution and can produce good enough results.

Let’s see how to build an ML system:

• Define the problem statement as clearly as possible and understand how it will benefit the company.
• Define success and criteria on which business can agree, it’s simple if the model does not help business, it will not be deployed to production and used.
• From the above diagram, you can see a significant amount of time is spent on data. The first step is collecting the data from different sources and the collection shall be in such a way that it can be done via an automated pipeline and shall be a scalable solution.
• Once data is collected next step would be checking and verifying the data through an automated pipeline itself on a set of methods.
• Understand how your model would be trained, does it needs to be trained in real-time? Based on the type of learning you want to incorporate for your model chose online or batch learning.
• Choosing the right model for your problem would be the next task. Finalize the model validate it with business criteria.
• Finally, deploy the model to production.
• Monitor its performance.

Models Of Machine Learning

Given below are the models of machine learning:

Well ML is vast and has many models but it can be broadly classified in the below sections:

1. Supervised Learning

• Classification:
  • • Logistic Regression
    • Support Vector Machine
    • Naïve Bayes
    • Decision Tree
    • Random Forest
    • Neural Network etc.
• Regression:
  • • Linear Regression
    • Decision Tree
    • Random Forest
    • Neural Networks etc.

2. Unsupervised Learning

• Clustering:
  • • K- Means Clustering
    • Hierarchical Clustering
• Dimensionality Reduction:
• • • PCA

Conclusion

In this article we have seen the ML system as a whole and what role does ML play in the whole system. We have also seen different components of the ML system and the steps involved in building it. We have also seen various models used in ML.

Recommended Articles

This is a guide to Machine Learning system. Here we discuss the introduction, how to build a machine learning system? and models respectively. You may also have a look at the following articles to learn more–

1. Machine Learning Life Cycle
2. Machine Learning Models
3. What is Machine Learning?
4. Introduction to Machine Learning