What is machine learning? By M Salim Bupati Rembang

Since the first time the computer was created, humans have thought about ways that computers can learn from experience. This was proven in 1952. Arthur Samuel created a program called game of checkers, on an IBM computer. The program can learn movements to win the game of checkers and save the movement into its memory. Then machine learning emerged.

The process of studying data
The term machine learning basically explains the computer process in studying data. Therefore, we will certainly continue to intersect with data, when studying machine learning. Data can be the same, however, the algorithms and approaches vary to get optimal results. Machine learning itself is one branch of artificial intelligence that addresses the development of data-based systems.

Follow the human way
Machine learning programs follow the way of human learning, which is learning from examples. Machine learning will study the patterns of the examples analyzed to determine the answers to the following questions.

Indeed, not all problems can be solved by machine learning programs. However, often complex algorithms can be easily solved by machine learning.

Some examples of machine learning-based programs used in everyday life are spam detectors, face detectors, product recommendations, virtual assistants, medical diagnostics, credit card fraud detectors, stock trading, customer segmentation, and automatic steering cars.

Ways of working
Basically, there are four types of machine learning work. The four are directed learning (supervised learning), non-directed learning (unsupervised learning), semi-supervised learning, and reinforcement learning.

Machine learning workflows include data collection, data exploration, model selection (linear regression, logistic regression, neural network, etc.), giving training to the chosen model, model evaluation, and prediction.

The initial accuracy of a machine learning program is usually very bad. Because at the beginning, the program "didn't know anything". However, over time, the more often we train programs, the more examples learned by the program. So, this program will be increasingly 'smart' and accurate.

For example, when we play the game Role Playing Game (RPG) that uses artificial intelligence. The first time we play with the RPG, then we will easily be able to win the game. However, after several games, the engine or game algorithm will learn from the previous patterns, so that it will be more difficult to beat.

What is your opinion? Apparently, the concept of machine learning is not as complicated as imagined, huh?