Machine Learning And Its Use Cases

What is Machine Learning?

Machine learning is not about a single factor that it can be defined by some words but in simple language, Machine learning is the process of equipping the computers with the ability to learn by using the data and experience like a human brain.

The main aim of machine learning is to create models which can train themselves to improve, perceive the complex patterns, and find solutions to the new problems by using the previous data.

The History

Machine learning is a science which was found and developed as a sub-field of artificial intelligence in the 1950s. The first steps of machine learning goes back to the 1950s but there were no significant researches and developments on this science. However, in the 1990s, the researches on this field restarted, developed and have reached to this day. It is a science that will improve more in the future. The reason behind this development is the difficulty of analyzing and processing the rapidly increasing data. Machine learning is based on the principle of finding the best model for the new data among the previous data thanks to this increasing data.

The Details.

In 1940s, based on the studies on the electrical crashes of the neurons, the scientists explained the decision-making mechanism of human by cannon and fire. In this way, the researches of the artificial intelligence started in the 1950s . In those years, Alan Turin executed the Turing Test in order to test the ability of a machine to imitate a human. The aim of the Turing Test was to measure the ability of the machine to make a contact with a human during an interview. If the machine performed worse than a human, it was successful. In 1956, the term ‘artificial intelligence’ was first used in a summer school held by Marvin Minsky from Massachusetts Institute of Technology, John McCarthy from Stanford University and Allen Newell and Herbert Simon from Carnegie-Mellon University. Until that time, Alan Turing’s term, ‘machine intelligence’, had been used.

In 1959, Arthur Samul created the checkers program, and then machine learning got its way. From those developments to the 1980s, there were some studies on abstract mind, information-based systems, which was called the ‘winter of artificial intelligence’. In the 1990s, artificial intelligence and machine learning studies accelerated due to the developments in game technologies. Nowadays, artificial intelligence and machine learning are used in lots of researches and work sectors.

The models.

• supervised learning
• Unsupervised learning
• Semi-supervised learning
• Reinforced learning

Supervised Learning: It is a method in which the present input data is used to reach the result set. There are two types of supervised learning: classification and regression supervised learning. Classification: Distributing the data into the categories defined on the data set according to their specific features.

Regression: Predicting or concluding the other features of the data based on its some available features.

Unsupervised Learning: The difference between the supervised and unsupervised learning is that in unsupervised learning the output data is not given. The learning process occurs by using the relations and connections between the data. Also, unsupervised learning doesn’t have a training data.There are also two types of unsupervised learning: clustering and association.

Clustering: Finding the groupings of data which are similar to each other when inherent groupings in the data is not known.

Association: Determining the relations and connections among the data in the same data set.

Semi-supervised Learning: supervised and unsupervised learning is inadequate when the labelled data are less than unlabelled data. In such cases, the unlabelled data, which are very inadequate, is used to deduce information about them. And, this method is called semi-supervised learning. The difference between the semi-supervised learning and the supervised learning is the labelled data set. In supervised learning, the labelled data are more than the data to be predicted. In contrast, in semi-supervised learning, the labelled data are less than the data to be predicted.

Reinforcement Learning: This is a kind of learning in which the agents learn via reward system. Although there is a start and finish points, the aim of the agent is to use the shortest and the correct ways to reach the goal. When the agent goes through the correct ways, s/he is given positive rewards. But the going through wrong ways means negative rewards. Learning occurs on the way to the goal.

Applications.

Image processing: In this method,it is aimed to process and improve recorded images. Some application areas where the image processor is used are as follows:

• Security systems
• Face detection
• Medicine (to diagnose diseased tissues and organs)
• Military (to process underwater and satellite images)
• Motion detection
• Object detection

Computational biology:

• DNA sequencing
• Finding a tumor
• Drug discovery

Natural language processing: It is aimed to investigate and analyse the structures of natural languages. It is possible to perform many applications with natural language processing:

• Automatic translation of written texts
• Question-answer machines
• Automatic summarization of text
• Understanding speech and command

Automotive, aviation and production:

• Detecting malfunctions before they occur
• Producing autonomous vehicles

Retail:

• Customized shelf analysis for persons
• Recommendation engines
• Material and stock estimates
• Purchasing -demand trends

Finance:

• Credit controls and risk assessments
• Algorithmic trading

Agriculture:

• Predicting yields or deficiencies by analysis of satellite images

Human Resources:

• Selecting the most successful candidate among a lot of applicants.

Energy:

• Calculating the heating and cooling loads for building designs
• Power usage analysis
• Smart network managements

Meteorology:

• Weather forecast via sensors

Health:

• Providing warning and diagnosis by analysing patient data
• Disease defining
• Health care analysis

Cyber security:

• Detecting the harmful network traffic
• Finding out address fraud

Thank You