Machine Learning: The Future of Efficient Governance
In next 5-10 years, we will see revolutionary transformation in every aspect of human life due to AI and it is said next world war won’t be nuclear but based on AI.
Photo Credit : ShutterStock,
With a deep focus on initiatives like ‘Digital India’, technology has become the nucleus of governance in India. The Modi Government has been making efforts to create an environment for digital growth and the rapid technological advancement that we are witnessing today, particularly in the field of artificial intelligence (AI), holds the opportunity to transform the way of governance in our country. Machine learning (ML), an application of AI that provides systems the ability to automatically learn and improve from experiences without being explicitly programmed, has emerged as a path breaking technology.
Having Aadhaar linked to pretty much everything in one’s life, government has already played a smart move collating data. ML & AI can help in achieving the scale and getting meaningful insights from this humongous data. By using ML algorithms like logistic regression, support vector machine or neural networks, one can create a statistical model which would show patterns emerging in the data and classifies them into different clusters. These models are then trained & tested with millions of historical examples for accuracy before being applied to new cases.
Governments in US, UK & other countries have started leveraging ML for efficient governance. They are using it for fraud detection, predictive analysis in healthcare domain, solving public transportation problem and even fight against terrorism and crime. In UK, the Ministry of Defense plans to extract critical information from terabytes of data it collects every day from images, speeches, sensors, radar, maps & vehicle data to stay one step ahead of terrorists and criminals. Closer home, Delhi Police recently announced that it will use AI to counter crime in the city, with the belief that AI-aided policing based on calculations and data analytics will help in efficient predictive policing.
ML in Finance sector
Income tax evasion is one of the biggest concerns not just for the government but all citizens who pay their taxes. When an Income Tax officer investigates a case, he analyzes the data for each individual manually going through hundreds of bank statements & investment documents to flag a fraudulent case. However, with the ML approach, we first collect data for historical fraudulent and non- fraudulent cases and build a model using an algorithm like logistic regression. A model is defined by features, which in this case could be sources of Income for an individual, taxes paid, number of properties & other assets owned, bank savings, cash transactions at banks/ATMs, credit card monthly expenses etc. This model will create a ‘decision boundary’ segregating fraudulent and non-fraudulent cases and now is capable to predict if a new case would be fraudulent or non-fraudulent, along with a level of accuracy. ML model “learns” from data and carves out its logic, without any rule or conditional logic. Once the model is built, it can identify such cases from thousands of them in matter of few seconds with greater accuracy! These new cases are fed back into the model to continuously refine and learn from new data.
In case of Bank NPAs, we gather data for the corporate loans given to companies in past. We build a model based on features like loan amount, type of industry, revenue & profit growth after loan, cash flows, existing assets & liabilities, number of repayments & defaults of interests in x years. Now when a new company approaches a bank for loan, it can run this new case through the model and predict chances of default. Banks can safeguard their interest with appropriate measures like higher interest or even penalties on default before approving it, or decide not to approve it at all. Today, without this prediction ability, it takes years before a bank realizes that a company would either go bankrupt or has to go through insolvency. Algorithms like neural networks are being used widely to predict bankruptcy in other countries.
Similarly, SEBI and regulators can leverage ML models to identify fraudulent activities like Insider trading, operators rigging, money laundering, investments via shell companies, tax evasion to name a few. For example, It is not difficult to identify a pattern in stock market transactions indicating insider trading around the time when company is about to announce some news. Many Investment banks and regulators in developed nations are already doing it using ML.
But how can ML help in controlling black money transactions? This could be tougher than other cases we discussed above. For one reason, there is not much digitized data recorded for these transactions. After all ML just provides a layer of intelligence on top of the data. Second, there are hundreds of illicit ways to hide black money or convert into white. But still few things can be looked at like Real estate deals, large cash transactions via banks and analyzing these against the sources of income. A model can be built which would reflect the correlation of income and reasonable assets one can own with that income. Anything which gets classified as outlier would be a case of investigation. We can identify lot of “benami” properties which are not linked to a reasonable source of Income.
What does future hold for ML?
The term ‘Machine Learning’ was coined as early as 1948 but its evolution & growth remained subdued because it needed enormous CPU power and memory for scientists to experiment. Today the challenge in ML is no longer hardware capabilities or complexity of algorithms but having the “relevant” data collected from various disparate systems for analysis and identifying the features to build the right model.
With most of the government departments digitized now, they all collect tons of data relevant to their business but they haven’t thought about how they can share necessary data with other government agencies in real time. Without real time interfaces between various entities, it would be impossible to achieve the desired results.
In next 5-10 years, we will see revolutionary transformation in every aspect of human life due to AI and it is said next world war won’t be nuclear but based on AI. UAE recently announced a separate minister for AI, aiming to make the government more efficient and streamlined by relying on AI technologies, while the kingdom of Saudi Arabia officially granted citizenship to a humanoid robot named Sophia.
The Indian government should partner with private players to overcome the challenges like skilled manpower in this era of digital transformation. India has no dearth of skilled technologists and scientists. Ministry of Electronics and Information Technology should also have a separate division which can embed technologies like ML & AI across processes at all government bodies.
With the right-intended policies & regulations for digital transformation, technology can surely help in achieving the government’s vision of ‘Maximum Governance, Minimum Government’. As we look towards the PM to bring structural transformation, I believe he is also looking at tech leaders & entrepreneurs in India to lead the technology change.
Disclaimer: The views expressed in the article above are those of the authors' and do not necessarily represent or reflect the views of this publishing house
Around The World