Machine Learning and Automation

A machine which has the ability to imitate the way of sensing things like humans and consequently has the ability to make deductions and communication falls under the umbrella of Artificial Intelligence. For example; a real time traffic tracking machine, that takes feed from the traffic camera, anticipating the congestion and alerts regarding the potential emergency situations makes use of the artificial intelligence. AI usually makes use of the Machine Learning. For instance, a machine can be taught to recognize the phenomenon using the statistical and mathematical tools, by
loading various numeric values, images and texts that correspond the phenomena to be learned in to an algorithm.

Machine Learning is a branch of AI, which is based on the idealization that machines are capable of learning and adapting through automation. It is a method of analyzing the data. Few of the applications of Machine Learning are:

1. The Google self-driving car.

2. Recommendation offers that come on Amazon or on Netflix.

3. Fraud spotting.

The process of making the hardware or software capable enough of doing actions automatically, is what we call Automation. Automation is something that is prevailing in the world and is present in every industry. Humans need to provide a set of instructions to the robots at first instance in order to make the process automated.

When the tasks of a Machine in the Machine Learning process are automated, they are called Automated Machine Learning (AML). Tasks like hyper parameter tuning, exploratory data analysis, model selection and pre-processing of data can be automated using Automated Machine Learning model.

Case Study:

At Airbnb, for creation of the customer lifetime value models (LTV) for their hosts and guests, machine learning is used. This framework helps them to improve their decisions at very granular level. The significant features of this model are; location and activity information, demographic information from their web and mobile devices. During the process of building the LTV model they came up with the algorithm called, Extreme
gradient boosted trees. Besides being biased, they fed the raw data to AML and came up with these findings:

The chart below shows the distribution of Root Mean Square Error.

The y axis represents the blueprints of the algorithms and the engineering measures.

The below chart gives the sense of what automated machine learning systems can be capable of.

Leave a Reply

Your email address will not be published. Required fields are marked *