How India-based Tookitaki Approach Predictive Modelling
The role of DSS has changed over the years, as it was primarily seen as a friendly interface to analyze raw data/documents, build scenarios and generate reports for businesses to support decisions. Scenarios or predictions were completely based on linear regression models and human analytical capabilities, which have seen revolutionary change as predictive models are created based on machine learning (ML) algorithms. Understanding this tectonic shift in trends, Singapore headquartered Tookitaki has built a platform based on artificial intelligence (AI) which can be termed as intelligent decision support system. These models use tree-based ML algorithms to provide businesses with higher accuracy and reinforce confidence in the decision-making process.
Tookitaki started as an advanced analytics engine to optimize online campaigns, especially on Facebook. “We were able to generate quick traction in Indian ecommerce industry, as our engine churned optimised results”, says Abhishek Chatterjee, Founder & CEO. The company operated mostly in prescriptive space, primarily to answer the ‘why’ behind good and poor performances of campaigns. With time, Tookitaki realised the potential of predictive power, implementing which, the engine has a bigger opportunity in data analytics space and in more matured markets like BFSI.
As organisations evolve towards making data-driven decisions, there would be a huge transformation from traditional Business Intelligence (BI) techniques to advanced analytics techniques, bolstering the growth in predictive analytics. The growth has been already noticed in developed western economies and is spreading towards early adopters in APAC regions including India. The growth of predictive analysis has also brought in a change in perspective for predictive analytics. For a considerable time, businesses were under the impression that predictive analytics was a consulting effort and performs ad hoc analysis on sample data using statistical packages like R, SaS etc. Now, platforms like TDSS are bringing a paradigm shift in such perspectives, as data scientists can launch and manage models on real data faster and even take them to production, without infrastructure and IT efforts.
What is TDSS?
AI based forecasting platform, named as Tookitaki Decision Support System (TDSS), breaks down the dilemma for data scientists to build accurate predictive models on real data and take them to production faster. Furthermore, these models come with in-built self-learning capability that auto-refreshes as new data gets added, thereby staying relevant to dynamic business environment.
TDSS lies in the intersection of data science and engineering, with primary focus on creating highly accurate models that can be deployed and managed in production environment faster and without any human intervention. There are three core features that differentiate TDSS from the available data science engines and traditional statistical packages: Feedback engine – in-built auto-refresh capability prevents model drifts and rather improves model performance over time, as new data gets added. Real-time predictions – automated production workflow helps extract predictions in real time from existing models via Rest API endpoints. Compute engine – pre-packaged compute and flexible storage layer allows deployment of models in 6 weeks and easy management of model environment.
How can TDSS benefit the Financial Institutions
KYC processes in banks are time and resource intensive efforts. TDSS offers an end-to-end KYC audit solution, which is much more advanced than the current rudimentary, deterministic and rule based platforms. “Our platform helps banks build a ML model that can accurately assign risk scores on new/existing KYC cases”, adds Abhishek.
The model auto-updates based on KYC policy requirements, resulting in improved precision in compliance to regulatory requirements without any manual complexity. It’s an appropriate solution for banks looking forward to have an AI-based internal KYC audit system to accurately detect mis-classified cases in significantly less time and without manual complexity, thereby making the process scalable and highly accurate.
Financial institutions need to optimize the monitoring of individual transactions and identify deliberate misconduct during the on-boarding and servicing of customers. The current models related to fraudulent behaviour transactions fall short of the self-learning nature and hence show significant drop in model accuracy as fresh data is added continuously, thereby making it irrelevant for business. The sparseness of fraud cases also prevents in building a highly accurate trained model. Tookitaki solution’s ability to handle the sparse data, build self-learning model and generate real time predictions makes it absolutely relevant and impactful for financial organisations looking to predict potential fraudulent behaviour accurately and instantaneously.
With solutions exploring a wide range of industries, Tookitaki intends to become a leading forecasting platform globally, with focus on applications around risk management. Currently, investing heavily on cutting-edge research around ML algorithms focussing on self-learning techniques and non-linear models, the company is all set to dive deep in providing a platform-powered approach to forecasting, rather than a consultative approach.