Supervised Learning: Train your Machine

What is Supervised Learning?

Supervised learning is a one of the most common type of machine learning algorithm. It is a two step process that involves: 

1.      Learning – Training a model using past data

2.      Testing – Assess that model accuracy by testing the model using unseen data

Let me explain the process with an example:

Observe the picture carefully. Can u help me segregate the items from the basket?

Now if you think about the process of segregation you will realize that your brain has been already trained. Based on the features like size, shape, color, texture and taste of the given fruits you could easily recognize and segregate them into different groups.

Similarly, in supervised learning machines are trained using the "labeled" historical data. These trained models also called as algorithms can predict and classify objects easily. Hence, we can say that our model has prior knowledge of what the output samples should be. Supervised learning models are fast and accurate as compared to unsupervised learning models that used unlabeled data and are untrained.

Types of supervised learning

Supervised learning broadly involves two learning approaches:

1. Regression
2. Classification

Regression:

Regression is mainly used in predicting, forecasting and finding relationship between two variables. Outputs of regression are real valued numbers that exist in continuous space. There are various types of regression that can be performed which includes linear regression, multiple regression and logistic regression. 

Examples:

• Predicting the probability and amount of fraud
• Predicting the price of the house
• Crime rate

Classification

Classification techniques are focused recognizing the pattern and predicting quantitative response. Hence, the output of classification algorithms falls into discrete categories. Classification algorithms include decision trees, K-Nearest Neighbor, Support vector machines and random forests. The above activity of fruit segregation can be considered as example of classification. 

Examples:

•  Fraud or no fraud
• Spam or no spam
• Credit- Good rating or bad rating
• Risk- high or low

Business Applications of Supervised Learning

Broadly, the use cases can be divided into 5 major areas:

1. Healthcare:

Healthcare has a wide range of use cases:

• Decision Support: Supervised learning based systems aid the radiologists in medical imaging recognition. The algorithms the medical image to offer the intelligible solution for detecting abnormalities across the body.
• Cancer detection and Prediction - Used to detect and classify tumors with higher accuracy level
• Drug Discovery: Predicting the molecular properties of the of the target molecule based on the training data set. Companies like Microsoft and BenevolentAI are using supervised learning. This will optimize manufacturing process, reduce the failure rate and eventually speed up the process thus saving costs.
• Insurance: Supervised learning methods are becoming extremely popular in the health insurance industry for predicting healthcare costs.

2. People Analytics:

Analytical tools are extensively used in decision making process especially when we need to manage employees. Managing employees means gathering data in a host of areas – employee attitudes and feelings, qualification verification, compensation management and list goes on. Here’s where the algorithms are used to analyze the wide range of data and generate smarter insights on:

• Performance measurement, retention and predicting who is on outward path
• Recruitment and workforce planning
• Detecting suspicious items in CV

3. Fintech:

The supervised learning combines programming, mathematics and statistics to provide insights on the structured and unstructured data. This improves the functionality of systems and markets.

• Predictions of stock markets
• Risk management
• Fraud Prevention

4. Information and Cyber Security:

Supervised learning algorithms are able to learn and gain knowledge of the internal and external vulnerabilities and do mapping against cyber attacks.

5. Marketing and Sales:

Almost all the major players across industries have online presence and use digital marketing. Supervised Learning is useful in recommendation systems and search history optimization to analyze user preferences and providing better customer experience.

Limitations of Supervised Learning

• The training data needs to be accurate as incorrect data will lead incorrect learning thus generating a loss
• Unwanted data will reduce accuracy
• The major challenge is pre-processing of data especially when dataset is quite huge
• Only labeled data can be used for supervised learning

Conclusion

We looked at the overall structure of how supervised learning exactly works through examples and diagrams. The applications of supervised learning vary across industries. It has helped business to optimize business process and save huge costs. Hence supervised learning will act as a low hanging fruit in data sciences for business.

References

1. Introduction to Machine Learning: Supervised and Unsupervised Learning | Analytics Steps. (2020). Retrieved 18 October 2020, from https://www.analyticssteps.com/blogs/introduction-machine-learning-supervised-and-unsupervised-learning
2. Guide, H., & Learning, 1. (2020). Pros And Cons Of Supervised Machine Learning | Pythonista Planet. Retrieved 19 October 2020, from https://www.pythonistaplanet.com/pros-and-cons-of-supervised-machine-learning/
3. Supervised Learning | What is, Types, Applications and Example | Edureka. (2020). Retrieved 19 October 2020, from https://www.edureka.co/blog/supervised-learning/#applications