Technology

One Step Towards Artificial Intelligence

The Chartered Accountant • October 2020 • pp. 82–87 (Journal pp. 445–450)

CA. Shubham Dosi

The author is a member of the Institute. He can be reached at shubham.mng@gmail.com and eboard@icai.in.

“The importance of Artificial Intelligence (AI) & Machine Learning (ML) in every sector including finance is in huge demand and increasing with every passing day. With every moment, technology is enlarging its scope with the intent to make business activities easy and creating global value as well. In return of the investments in AI & ML, we get multiple times excess in the form of permanent cost reduction, systematic resource optimization, improved efficiency & productivity and so on result into creating a global value.

Read on to know more on complete discussion mainly focusing on how CA professionals have extreme qualities and opportunities in current “tech-era” to build values which make differentiation among other professionals.”

Artificial Intelligence (AI) has grown to become an intrinsic part of the world’s Economy. In a broader sense, we can say that Machine Learning (ML) and Deep Learning (DL) are adjacent to Artificial Intelligence. In this ongoing “FIN-TECH” era, everyone looks forward to achieve financial synergy and build vision to accomplish the value of data on which whole globe can be dependent. ML in finance is restructuring the entire service industry in the past 2 decades. Artificial Intelligence is a technique of developing brilliant machine programs which can think much better than humans because of its diverse training input data. Machine learning is an approach to examine the machine program to constantly improve the efficiency and accuracy of such machine program. In this article the primary object is to discover opportunities with Fin-Tech methodology to grow our CA profession and contribute in nation building. The article is divided into two major categories, ‘Overview of AI & ML’ and ‘What we can do and how we can enhance the worth worldwide?’

A. Overview of Artificial Intelligence and Machine Learning

India has become a prime source destination across the world to establish both internal as well external IT infrastructure. Additionally, India is progressing to provide the best financial service with IT collaboration. Machine Learning has been proven as a revolution in technology in the past few years. Data analytics have now become a preference for any organisation to capture market share and enhance the intrinsic value as such. Therefore, it can be easily said that both AI & ML have been creating correlation within each other and focusing on “Learn with Logic” concept.

AI is about intellect with reasoning, problem solving and learning capabilities, similarly ML trains with several historical data and behaviour and has become competent to predict and forecast the result. AI & ML can efficiently perform multiple tasks at a time like data visualisation, forecasting & predictions, statistical analysis, translation, resolving puzzles, arithmetic or non-arithmetical computation in a clever manner. Machine Learning is also required to build an algorithm, depend on every incident or diverse instances which assist to train the module.

Classification of Machine Learning

Machine learning is classified into three comprehensive categories:

1. Supervised Learning

Task-driven algorithm / Historical train data

2. Unsupervised Learning

Data-driven technique / Cluster analysis

3. Reinforcement Learning

Learn by their own / Reward & feedback system

Now we are going forward to understand the concept of all these categories in phased manner:

1. Supervised Learning

Supervised Learning is one of the most primary type learning in which Learning comes through historical events, historical data or original dataset which is ordinarily called as “Train Data”. Supervised Learning is one of the utmost robust programming techniques which has been proven at various circumstances. In the first phase the algorithm has been trained by providing various data at large size including diverse data in a form of Train Data with help of that program can understand logic, relationship in various parameter of the data, understand the problem and continuously focus on improving results. In a Train Data, user provides multiple parameters based on which algorithm becomes effective and much more capable to predict accurate results on “Test Data” or on future events. The algorithm is much more capable to build prose and cause relationship between structured parameters. Supervised learning makes a machine emphatic and capable to learn with logical experience which helps to produce reliable predictions. We can hence make a favourable remark by saying, “Supervised Learning is a task driven program or task driven algorithm technique”. Supervised Learning is famous to resolve regression & classification complicity.

To review the final outcome and accuracy of test data, a user always computes result from test data and compares with original results. With the test outcome, the user can always focus on continuous improvement and making crucial changes in algorithm to enhance the accuracy and reliability of the data. Regression also can be divided into 6 extensive categories:

  1. Simple Linear Regression
  2. Multiple Linear Regression
  3. Polynomial regression
  4. Support Vector Regression
  5. Decision Tree Regression
  6. Random Forest Regression

2. Unsupervised Learning

Unsupervised Learning has an additional attribute, which is learning without training data. As opposed to Supervised Learning, Unsupervised Learning is a data driven technique. In the unsupervised learning program human intervention is not much required to train the program. Machine algorithm understands the data, identifies patterns, identifies structures of data by their own explicit hidden attributes. Unsupervised Learning primarily focusses on cluster analysis which is helpful for data analysis to identify hidden relationship in various parameters to identify pattern or classification of the data. Unsupervised learning is practically not relevant for predictions or forecasting, this method is mainly useful to sequence analysis or pattern mining in the dataset. Common clustering algorithms include the following:

  1. K-Means clustering
  2. Hierarchical clustering
  3. Gaussian mixture models
  4. Self-organising maps
  5. Hidden Markov models

3. Reinforcement Learning

Reinforcement Learning is an approach to AI and mainly known for the distinctive quality called “Learn by their Own”. It means that machine algorithm or program is much capable to learn with data as similar to how human beings learn in their lives with experience. Its programs or algorithm attributes always improve by itself and learn from diverse situations with various methods. Every successful output is reinforced and every unsuccessful output again improves the logic and makes adequate changes for improving the subsequent output. In simple terms, learning would yield both positive as well negative reinforcement. In case the algorithm finds the accurate and appropriate result, the interpreter reinforces the solution by providing rewards to such algorithm program. The intention behind the reward is to improve the effectiveness or accuracy of the output.

Deep Learning

Machine Learning also includes Deep Learning, a specialised direction that carries the future of Artificial Intelligence and ensures the success of Machine Learning as well. Deep Learning eminent “Neural-Networks” is a kind of programming or algorithm that was found to look like the physical form of a human brain. Worldwide Neural networks seem to be the most effective and efficient way for Artificial Intelligence research hereafter. In Deep Learning we can build a relationship between more complex parameters in a very effective manner. In Deep Learning, input data processes within multiple layers and after that, algorithm provides logical and valuable results. Deep Learning algorithm has the ability to learn from unstructured unlabeled data and in this way, it creates a different quality with Machine Learning.

Moving forward to discuss about What actually “Neural-Network” is and How effectively it works?

As in our brain neurons are the key players to process and manage the information, similarly in Deep Learning we create an “Artificial Neuron Network” (ANN) to process input data and produce accurate result for complex data with the assistance of Machine Learning or algorithm. Neural-Network is a concept in which the user develops an Artificial Neuron Network with algorithm or program, thereby paving way for it to work like the human brain.

In our brain more than 100 billion neurons are working to manage our whole-body including the mind. In Artificial Neuron Network, organic neurons have to be substituted by multiple mathematical functions. There are lots of neurons in an artificial network, each with a unique and essential function that assorts the data given to the program. This artificial neuron constructs in various layers to produce accurate results with proper interpretation and understanding of the data. This various layers in neuron are also called as nodes and assign a weight which work as filter for processing the data. All the layers are divided into 3 categories, namely Input layer output layer and hidden layers which are the integral parts. Neural Network is also categorised into three broad terms:

  1. Recurrent Neural Networks
  2. Convolutional Neural Network
  3. Multilayer Perceptron

Right now, we are not going to discuss all the three in detail because we are much more interested to discuss about the beauty of all such networks and how it will be better in our profession and how we can create synergy in our profession.

B. How AI & ML be succeeding path for our Professionals

Artificial Intelligence (AI) & Machine Learning (ML) has become more popular due to easy availability of vast quantum of data at affordable costs. AI & ML have been reshaping the whole financial industry over the past few years. It is difficult to determine the future of financial services without AI & ML. This article is written to give more impetus on AI & ML to provide a capacity to build global value of our services and qualities thereof; how one can set his/her profession as a benchmark worldwide. This article would touch upon the importance of the said revolution in technology, understanding and preparing to bridge the gap between profession and advancement in technology.

While we are on the subject of how IT creates revolution in every segment and how revolutionary measures and techniques are followed everywhere, finance has also simultaneously been enhancing to value to match with IT to provide excellent services. No one can deny that AI & ML have created a situation of panic around the professional community and created a hype that AI & ML will drastically impact the profession, but at the same time AI & ML have opened a diverse path to make career and derive the opportunities to succeed. We need to take just one step forward to move towards IT Collaboration in Finance. In the upcoming discussion, let us analyse the road map to match an individual’s profession with AI & ML era:

  • Financial Planning & Analysis (FP&A): When we work in Finance Planning & Analysis (FP&A), with the help of AI & ML we can analyse and define the data from scratch to extreme level for principal conclusion. We can recognise multiple correlation between different parameters of the data and can parallelly build a relationship in various inputs with data models used in AI & ML in the dataset which is primarily useful for taking strategic decisions. AI & ML have been proven as best approaches to forecast and predict a much more accurate result and output with various parameters in dataset. It helps to correlate both internal & external and micro & macro factors which are directly or indirectly relevant for analysis. With AI & ML we can identify the worth of even the smallest data in the dataset and can be able to understand the contribution in analysing whole data. Ordinarily organisation ignore those factors which directly not related to corporate decision making, but with AI/ ML tools, organisation consider all the factors even having less relevance which directly or indirectly worthful to make a strategic decision. Will take an illustration to understand in broader way how AI & ML proven as a path of success in FP&A. It is difficult to understand the human behaviour but with help of past dataset of customer’s age, location, gender, their qualification, their money power and their product selection we can make a various permutation and combination to anticipate will customer be associate with us in upcoming period or customer will get separate from us. We have to make a strategic decision to understand is there any requirement to make changes/ upgradation in quality of our product and service or is their any obligation to modify price of our product and services respectively.
  • Automation of Repetitive Operations: Automation is one of the most common requirements in every segment and in every organisation to improve the accuracy of result, enhance the efficiency of personnel, scale up the quality of product or service, match ourselves with global standards, systematic cost reduction and optimise resource. We can build an algorithm to reduce repetitive activity with successful AI & ML module to improve productivity which is useful to enhance productivity not only in a single department but in the whole organisation. A small illustration of this automation is sending invoice directly to customers with AI & ML technique without human intervention. To determine credit worthiness or credit rating of any consumer, we can train the data with thousands of entries which includes numerous combinations in dataset through which algorithm can make maximum permutations and combinations to derive the most accurate result of test data, it is a very complex task to derive credit value of millions of persons simultaneously, and this is where AI & ML can be of great help for an organisation.
  • Portfolio Management & Robotic Advisory: Recognise intelligent behaviour in machines to manage portfolio and provide choice to view and analyse the return on their own fund with addition impart the recommendation to build robust portfolio with diverse segments. Professionally we can engage in financial services where the primary focus is to offer solutions to Individuals, Institutions or Corporate entities and assist them in exploiting their excessive funds to yield capital. Nowadays, the world is moving towards a robotic advisory to build a faultless service so as to keep their service as a benchmark in the respective industry. Calculation of financial statistics such as daily return, volatility, cumulative return on portfolio, optimise portfolio allocation, computation financial strength of company, time series etc. are few illustrations of Fintech services. Building thematic investment strategies and robust portfolio with study algorithm, computation of NPV, making Capital Asset Pricing Model (CAPM). It is true that the cost of service provided through AI & ML is much cheaper than consulting as a human financial consultant. With AI & ML, it is possible to provide virtual assistance to our customers anytime and anywhere.
  • Algorithmic Trading & Financial Market Modelling: Build a strong algorithm through which user can independently understand the movement in financial market. Data module or algorithm always keeps client ready to instantly respond on real time challenges. He himself can make a strategic portfolio based on present scenario and market conditions. Investor can be making a choice of selection based on various investment strategies in high-frequency trading environment which includes quantum strategy too. Predicting the future trend, movement and price based on historical trend with AI &ML or technical statics, risk of loss in financial market have also been reduced to a great extent. Professionals are playing vital role to make strategies for algorithm trading (algo-trading) which are very much prominent worldwide. We can make an advanced screening tool with various author’s strategy like Value Investing, Growth Investing, CANSLIM, The Naked Traders etc. User can independently select the strategy and explore estimated future outcomes of their particular portfolio.
  • Audit Risk Reduction & Robotic Computer Audit Technique (RCAT): AI & ML have the potential to analyse data and provide a more relevant observation at the time of Audit. Auditor can reduce their risk level at the time of audit. Auditor can check entire entries with AI & ML algorithm or programming which not required to set sample for audit to accompany with this entire verification auditor is capable to reduce control risk and detection risk at negligible level and also can make sure the inherent risk gets reduced as compared to pre-AI & ML era and can make a true & fair audit opinion in the report. In any organisation if all these three risk are under control or at reduced level then we can say the risk management of the particular organisations are well governed and properly managed. Directly or Indirectly it increases the confidence of all the stakeholders of such organisation. Auditor can use AI & ML as a Robotic Computer Audit Technique (RCAT). With AI & ML auditor can identify any exceptional or suspicious transaction that have taken place in the company during the audit period. At an advanced level, the auditor can verify the pattern of transaction, transaction price comparison for every scenario, quantity and price relation for every transaction to identify whether the transaction is normal or not. Auditor role is to now monitor the procedure, interpretation of data, and to ensure effectiveness and efficiency of data module in AI & ML.
  • Decision Support Services & Internal Controls: AI & ML is a replacement of professionals. Simultaneously, it is an opportunity to provide much better decision support services with IT collaboration. We can understand the relevance of data in depth and can make more strategic decisions which would absolutely be worthful for organisation. With AI & ML perfect governance structure can be set-up and any risk associated with internal control can be easily identified and managed accordingly. Professionals need to focus on how machine learning can be leveraged to facilitate our roles & responsibility towards stakeholders. With AI & ML professionals can allocate their valuable time to improve efficiency & productivity of services with the mission to increase the market size.
  • Big Data Analytics Across Business Functions: Big data analytics is becoming a prime choice for various companies and for that they are looking to such professionals who can build logic and can summarise the logic into algorithm. Not only pertains to finance but also having a choice in diversified streams like supply chain management, product management, sustainable market selection, valuation etc. which makes it easy to standardise the global value of any organisation. Most companies are focusing on data analysis through application of multiple statistics on whole data and visualise the entire data.
  • Fraud Detection & Risk Mitigation: Technology has been playing an integral role in many phases of the financial ecosystem. High volume of historical data is now easily available at moderate cost with that it’s very feasible to examine entire data with extreme logical concepts. Fraud detection and risk mitigation are also now easy to control. To verify large number of data entries it’s easy to observe fraudulent or suspicious arrangements like money laundering, unfair transactions, etc. To determine the real long-term value of any entity, technology has been an integral part of it and based on techno level anyone can determine the value of it.
  • Deep Learning in Document Analysis: Documentation and the critical analysis thereof also a complicated job for anyone. Very instant advantage of Deep Learning has transformed image recognition accuracy beyond our capability. Document analysis is also an ideal illustration of AI & ML in finance sector. To reduce human efforts and simultaneously increase efficiency with accuracy with cost optimisation, Machine Learning is seen as a diamond in a coal mine.