By Gokul Siddharthan J, DCMME Graduate Student Assistant

Machine Learning

Machine learning and Artificial Intelligence are part of the same family. In fact, machine learning is a branch of AI-based computer systems that can learn from data, identify patterns, and make decisions without human intervention. When exposed to new data, computer systems can learn, grow, change and develop themselves.

Machine learning and AI is everywhere. There is a high possibility you are using it and don’t even know about it. Some of the instances where machine learning is applied are Google’s self-driving car, fraud detection, online recommendations such as in Amazon, Facebook, Google Ads, Netflix recommendations, and many more. Traditional data analysis was done by trial and error, but this approach isn’t feasible when data becomes large and heterogeneous. Machine learning proposes smart alternatives to analyzing huge volumes of data through fast and efficient algorithms and analysis of real-time data. It is able to produce accurate results and analysis. Other major uses of machine learning are in virtual personal assistants, such as Alexa, Google Home, Siri, in online customer support, where chatbots interact with the customer to present information from the website, in predictions while commuting, Traffic predictions like Google Maps and online transportation networks like Uber, in social media services, such as Facebook’s people you may know, face recognition of uploaded photos. These are a few examples, but there are numerous uses where machine learning has been proving its potential.

So how do machines learn? There are two popular methods, supervised learning and unsupervised learning. About 70 per cent of machine learning is supervised, while unsupervised is around 10-20 per cent. Other less often used methods are semi-supervised and reinforcement learnings. In supervised learning, inputs and outputs are clearly identified, and algorithms are trained using labelled examples. The algorithm receives inputs along with a correct output to find errors. Supervised learning is used in applications where data predict future events, such as fraudulent credit card transactions. Unlike supervised learning, unsupervised learning uses data sets without historical data. It explores surpassed data to find the structure. It works best for transactional data, i.e. in identifying customer segments with certain attributes. Other areas where unsupervised learning is used are online recommendations, identifying data outliers, self-organizing maps.

Google’s chief economist Hal Varian adds, “just as mass production changed the way products were assembled, and continuous improvement changed how manufacturing was done, so continuous experimentation will improve the way we optimize business processes in our organizations.” It’s clear that machine learning is here to stay.