Hemal Shah, Executive Director, Product Delivery & Regional CIO APJ, DELL
With the progress made in internet technology and ubiquitous smart devices, there has been exponential growth in the data that is generated today. To put this in perspective, imagine the kind of data that is generated via computers, servers, social media and the smart devices by about half of the population in the world. In 1995, less than one-percent of the population had access to the internet and today about forty-percent of the world's population has access to it and the number of users continue to grow. This provides enterprises with both a unique challenge and an opportunity to meaningfully use the generated data.
Historically, reporting and data mining techniques were used to build intelligence on the data. However, with the rapid growth of technology, there are improved intelligent algorithm-based tools that can be put to good use to understand the data and use it for proactive and predictive decisions. In many of the use cases which require various sparse decision points and cannot be predicted, the algorithms have to be coupled with the human intelligence factor in order to narrow down the decision trees. For these scenarios, the use of predictive computational techniques pose a challenge due to the explosion of data, dependencies on human intelligence and risk of missing paths. For these problems, it is fitting for one’s machines or compute devices or programs to adapt, earn the problems and gather intelligence as it matures.
The solutions lies in Machine learning - a computing technique that uses data and statistical analysis to extract knowledge from the data and act on the information learned from the data. The available data is too complex and unstructured, so machine learning develops algorithms to discover knowledge, based on statistical and computational principles. Machine learning is usually classified into supervised and un-supervised learning. Supervised Learning is training the machine to infer from labelled data. The common techniques used here are Bayesian networks, expert systems and neural networks. Un-Supervised learning is teaching the machine to infer from a set of unlabelled data. The most common among this is cluster analysis for finding hidden patterns and grouping data, and techniques like Kohonen maps.
Machine Learning has been around for a while in different guises and has gone through phases of evolution from the mid-50s. ELIZA, Artificial Neural Networks, Statistical Artificial Intelligence in the late 19th century and Big data science present today are all Machine learning techniques that are fundamentally focused on meaningful use of data. Many of the areas are currently using different machine learning techniques. Key examples include financial prediction, weather analysis, spotting customer buying patterns, improving web searches and pattern recognition, among others.
At Dell, we are using machine learning techniques to understand social patterns from customers - including buying patterns to upsell, improving the order cycle times by analyzing the supply chain, predicting sales and also improving scalability and resilience of our applications, and continually growing areas where we can apply intelligence to solve business problems. We are constantly encouraging our team members to innovate and look for opportunities that provide great value; machine learning and big data are big part of that.
However, with the data explosion and penetration of computing devices and sensors – there is a big increase in complexity and a unique opportunity for harvesting and making programs more intelligent. As an IT organization, machine learning can be used to solve a number of real life problems for businesses that have been using traditional methods. This could include the ability to spot buying patterns amongst targeted customers; optimizing the supply chain for reduced costs; and product lifecycle planning, etc. In the context of impact on the day-to-day lives of consumers, Machine Learning has the potential to improve web searches by using data driven algorithms; analyse data for predicting and preventing genetic diseases; reduce traffic congestions in large cities around the world; improve financial services, predict natural disasters; enable voice and face recognition and make way for self-driving accident free automobiles, etc.
Machine Learning, from a discipline of a handful of computer scientists a few years back has grown to a mainstream knowledge area today, across verticals. This needs to be an integral part of IT, and in the upcoming years Machine Learning will continue to play an important role in data analysis and the application areas.