Machine Learning (ML) can be a hard concept to grasp because there is so much to it.
Typical coding is done through algorithms where the input plus the program equals the output. However, in ML, only the input and output are given, and the program must be created from them.
There are also different types of ML: supervised, unsupervised, and semi-supervised.
1. Supervised Machine Learning
In supervised machine learning, the data is attached to labels in order to train it. Once the data is trained, the testing would be done to then attach those labels based off of the testing data.
Supervised ML would be used in classification. Whether it be classifying sports based off of keywords or classifying if someone has heart disease, supervised ML is used.
2. Unsupervised Machine Learning
Unsupervised machine learning is completed by using data without labels (in other words, there are no comparisons done between data).
To do this, clustering is often used to find patterns. When some of the data have a bunch of similar cells from the features, they are often clustered together.
There are different types of clustering as well. K-means clustering will separate the clusters based off of which other records each record is closest to.
DBSCAN clustering is used for clusters that contain the same density. If the clusters look like parabolas or other shapes, then DBSCAN clustering is used.
Supervised vs. Unsupervised ML
From: Unsupervised vs. Supervised Photo
3. Semi-supervised Machine Learning
Semi-supervised machine learning is basically just a combination of supervised and unsupervised learning. Some labeled data is used at the beginning to start and then most of the data is used without labels to then create the model.
Which to Use
Machine learning has many different techniques. To determine which method is the best to use, look at the data. Exploring the data will help decide which method is the best and will be the most accurate.
ML is necessary to be a successful Data Scientist and can help solve real issues within businesses today.
To learn more, please visit: What is Machine Learning? | IBM