• BRIEF MATERIALS - FINANCIAL CLUB BREAKFAST DIALOGUE

BRIEF MATERIALS - FINANCIAL CLUB BREAKFAST DIALOGUE

Date: 24 April 2019
Time: 08:00 - 10:00
Financial Club Jakarta
Graha CIMB Niaga 27th floor Jl. Jenderal Sudirman Kav. 58
Jakarta  
get directions

  • Summary

A. Why we should consider Machine Learning and Artificial Intelligence in finance?

Even though the challenge to use Ml and AI is quite big, many financial companies have used this technology. The figure below shows that financial service executives consider machine learning seriously, and they do it for many reasons:

• Reducing operational costs by automation process.

• Increasing revenue by better productivity and improved user experience.

• Better compliance and strengthened security.

Currently there are various kinds of learning algorithms with open-source devices and tools that are very compatible with financial data. In addition, established financial services companies have large funds that they can spend to provide sophisticated computer hardware.

Because of the quantitative nature of the financial domain and the large volume of historical data, machine learning is ready to be used to improve many improvements to aspects of the financial ecosystem.

That is why so many financial companies invest heavily in R & D using machine learning. While a small portion still considers it expensive and tends to ignore the AI and ML.

B. In what cases can Machine Learning be applied in finance?

Let's take a look at some promising Machine Learning applications in finance.

B-1. Automation Process

The process of automation is one of the most common applications of Machine Learning and AI in finance. This technology makes it possible to replace manual jobs, automate repetitive tasks and increase productivity significantly.

As a result, Machine Learning enables companies to optimise costs, improve customer experience, and improve services. Here is a case of using machine learning automation in finance:

• Application of Chatbots technology

• Call Centre automation.

• Document automation.

• Gamification of employee training and many more.

Here are some examples of process automation in banking:

• JPMorgan Chase launched the Contract Intelligence (COiN) platform that utilises Natural Language Processing, one of the Machine Learning techniques that process data text. This solution processes legal documents and extracts important data from the data they have. Manually reviewing 12,000 annual commercial credit agreements usually takes around 360,000 working hours. In fact, Machine Learning makes it possible to review the same number of contracts in just a few hours.

• BNY Mello integrates automated processes into their banking ecosystem. This innovation is responsible for an annual savings of $ 300,000 and has resulted in various operational performance improvements.

• Wells Fargo uses AI-driven chatbots via the Facebook Messenger platform to communicate with users and provide assistance with passwords and encrypted accounts.

• Privatbank is a Ukrainian bank that implements chatbots assistants on all cellular and web platforms. Chatbots speed up the resolution of general customer requests and make it possible to reduce the amount of human power assistance.

• The use of e-money on toll roads in Indonesia.

B-2. Security

The threat of security in the financial sector today is increasing along with the increasing number of third party transactions, users and integration. The Ml and AI algorithms are very good at detecting Fraud.

For example, banks can use this technology to monitor thousands of transaction parameters for each account in real time. The algorithm checks every action taken by the card holder and assesses whether an activity is matched with the characteristic of the user. Such models show high-precision fraud behaviour.

If the system identifies suspicious account behaviour, it can request additional identification from the user to validate the transaction, or even block transactions altogether if there is at least 95% chance it is fraud. The ML-AI algorithm only needs a few seconds (or even seconds) to assess transactions. Its speed in helping to prevent fraud in real time is very reliable, not only finding after a crime has been committed or has occurred.

Financial monitoring is a case of other security uses for ML / AI in finance. The Data Scientists can train the system to detect a large number of micro payments and mark money laundering techniques such as smurfing with available data.

The ML-AI algorithm can also significantly improve network security. The Data Scientist can train the system to recognise and isolate cyber threats, because ML-AL is a tool that has no capability in analysing thousands of parameters and real time. And, the possibility of this technology will strengthen the most sophisticated cybersecurity network in the near future.

Adyen, Payoneer, Paypal, Stripe, and Skrill are some of the leading fintech companies that invest heavily in ML-AI security.

B-3. Underwriting/Credit Scoring

The ML-AI algorithm is very suitable for guarantee tasks that are very common in finance and insurance.

The Data Scientist trains models on thousands of customer profiles with hundreds of data entries for each customer. A trained system can then carry out the same credit guarantee and assessment tasks in the real-life environment. Such a scoring machine helps human employees work faster and more accurately.

Banks and insurance companies have a large amount of historical consumer data, so they can use these data to train ML models, or they can utilise data sets produced by telecommunications companies or have large utilities.

For example, BBVA Bancomer is working with an alternative credit generating platform, Destacame. The bank aims to increase access to credit for customers with a history of microcredit in Latin America. Destacame accesses bill payment information from utility companies through an open API. Using the bill payment behaviour, Destacame generates a credit score for the customer and sends the results to the bank. BDO has experience in creating Credit Scoring for Jamkrindo which is intended for MSMEs in Indonesia.

 

B-4. Trading Algorithm

In algorithmic trading, ML and AI help us make decisions on better trade transactions. Mathematical models monitor news and trading results in real-time and detects patterns that can predict stock prices, up or down, accurately. Then they can act proactively to sell, hold, or buy shares as predicted.

The ML-AI algorithm can analyse thousands of data sources simultaneously, something that cannot be done by human traders.

The ML-AI algorithm helps traders get slight profits above the market average. And, given the large volume of trading operations, that small profit often accumulates into significant profits in the end.

B-5. Robo-Advisors

Robotic advisors or Robo-advisors are now commonly used in the financial sector. Currently, there are two main ML applications in the advisory / advisory domain.

Portfolio management is an online wealth management service that uses algorithms and statistics to allocate, manage and optimise client assets. Users enter their current assets and financial goals, for example, save one million dollars at the age of 50.

A robo-advisor then allocates the current asset across investment opportunities based on the risk preferences and desired objectives.

Many online insurance services use robo-advisor to recommend personalised insurance packages to certain users. The customer chooses a robot advisor rather than a personal financial advisor because of lower costs, as well as personalised and calibrated recommendations.

C. The Key to Utilising ML-AI in Finance

• Financial companies mostly use Ml-AI for process automation and security.

• Before collecting data, you must have a clear view of the results you expect from Data Science. There is a need to establish appropriate KPIs and make realistic estimates before the project starts.

• Many financial service companies require data engineering, statistics, and data visualisation through Data Science and ML-AI.

• Larger and cleaner training datasets with good quality will produce more accurate predictions to produce the best ML-AI solution.

• You can retrain your model as often as you need without stopping the ML-AI algorithm.

• There is no universal ML-AI solution to be applied to different business cases. ML-AI is very specific and unique for each problem and different data sets.

• R & D in ML-AI is still considered expensive.

• Giant technology corporation like Google are creating latest ML-AI solution. If your project concerns use cases as they have developed, then you cannot expect to surpass the algorithms developed by Google, Amazon, or IBM.