Petabytes of data are being generated with every interaction between Man and Machine, and more so with Machine to Machine interactions. This data captures the fine print of the sequence of events that have occurred before and after an action. The actions may have a success or failure associated to it. The ability to learn from the past failures and build on the strengths of past success differentiates the organization’s ability to be a market leader. With every passing day, complexity of data is continuously posing a challenge, with exponential increase in the data volumes, diversity of data formats and explosion of unstructured data like video, voice and images, and the rate/velocity at which it is getting generated. Sophisticated Location Aware Mobile technologies that give significant processing power on the move, and widespread acceptability of social media platforms have resulted into huge amounts of information-rich data. This also opens up interesting opportunities for building technologies that support and provide capabilities to harness the benefits of information/knowledge that is embedded in data.
The huge amount of Information rich data complements the business with non-obvious actionable insights. Visualization of large amounts of data with automatic knowledge discovery is a great value proposition that is available as an opportunity, and Predictive Analytics provides the capability to find actionable information which has substantial positive impact in the future. Google, Facebook and many other successful initiatives focused around analytics on top of big data have provided compelling evidence of how predictive analytics over big data can make billions for the enterprise.
Predictive Analytics has it applications in various domains; however the prominent use case we see is in marketing vertical for retail, ecommerce or m-commerce domains. In addition to data that is generated within the enterprise, information available on forums like Blogs, Tweets and Facebook likes/comments are used to get a 360 degree view of the customer. Enterprises that have integrated marketing strategy with predictive analytics embedded into the operational data, have significantly improved customer experiences due to the ability to anticipate the needs of the customer, and estimate the probability of the next best action of the customer on real time basis. This enhances the enterprises ability to optimize and automate decisions to better manage the needs of the customer and increase the chances of success in future.
2014 is going to be a year which will be concentrated around the power of data. Increasing number of companies will be inclined to look for analytics driven business insights, and would want to include predictive analytics into every operations that influences the end user, may it be the customer, client or employee. Knowledge is power, and data has knowledge rooted in it, and hence we will find organizations making bigger investments in Big Data Technologies, Predictive Analytics, Data Privacy and Data Security.
Predictive Analytics has some challenges as well. Predictive Models are trained on historical data, and large volumes of data need to be crunched, increasing the time and space complexity while training the model. Once the model is trained then it must be possible to extract the model information and deploy it as a part of the operational system for predictions on real time data. The concept that is captured by the trained Predictive model may drift over time which demands the necessity to retrain the model automatically. Hence, technologies that enable predictive modeling on big data streams and provide predictive insights on real time data, besides having the ability to capture concept drift automatically and trigger the regeneration/retraining of models is going to be the flavor of the 2014.
The need for highly skilled professional is a must for Predictive Modelling and this skill set is hard to find. This poses another interesting challenge for making predictive analytics accessible to the enterprises that do not have such a resource. The ability to abstract the complexities of model generation and create an interface for the end users (at different levels of predictive understanding), enabling them to use analytics as a basic utility service in their daily activities will be in the forefront of new initiatives. Analytics as Service is going to be a prominent theme in 2014. Huge investment/ initiatives in the Services industry around Centre of Excellence for Big Data Analytics will also be visible.
The Internet of Things (IoT) and technology landscape enabling it is expected to be $8.9 Trillion market in 2020, according to IDC. This leads one to imagine the awe-inspiring impact this will have in the world around us. Smart houses, offices and towns, every appliance in this context will have the ability to monitor and capture every delta change around it, report status and respond to requests from other devices located at remote locations. This completely changes our world view of the things around us. With the beginning of this new era we will see an increase in initiatives concentrated around Predictive Analytics, either building solutions for specific industry verticals, or building apps around the Internet of Things.
This also necessitates a fresh look at research algorithms in the predictive domain. Single pass, simple and lightning fast algorithms which scales with big data and optimally uses the computational technologies will also be the flavor of this year.
Dr Pallath Paul, director, Technology and Innovation Platform Group, SAP Labs India
