Machine Learning in Finance

An Introduction To Machine Learning

Machine learning is a branch of AI which aims to give machine the ability to learn without being explicitly programed. By feeding algorithms enough data, we enable machines to learn solve sophisticated tasks without knowing exactly what the problem is or how to fix it. The aim of ML is to solve tasks in a particular discipline with minimum knowledge in the corresponding area. Therefore, machine learning is implemented to various subjects, and it is not likely to decrease in popularity. Depending on what problem(data) they try to solve, machine learning algorithms are segmented into three groups.

1. Supervised Learning

Supervised learning is the most common type of learning that apply what has been learned in the past to new data using labeled examples to predict future events. Based on the type of output data (continuous or discrete), different supervised learning algorithms are implemented.

• Regression
  In this type of learning, machines tend to predict or approximate what the value of an output is, based on the given input data. For example, predicting the size of a house given its size.
• Classification
  Classification algorithms assign a number of observation to a set of categories. For instance, we might want to predict that if a tumor is benign or malignant based on the size of the tumor and the age of the patient

2. Unsupervised Learning

Unlike supervised learning, unsupervised learning derives solutions to problems corresponding to unlabeled data. The application of unsupervised learning are more diverse compared to that of the supervised learning. Depending on the algorithm, they adress dimensionality reduction, clustering, anomaly detection, and etc. Here the aim is not to predict outputs, but rather to make inferences from hidden structures in our data.

3. Reinforcement learning

This learning method interacts with its environment and improves by learning from errors and rewards. Trial and error search and delayed reward are the most relevant characteristics of reinforcement learning.

ML applications in Finance

Today, machine learning plays an integral role in many phases of the financial ecosystem. Given the fact that machine learning is a very broad concept, we will learn how some of the machine learning methods are applicable to finance.

• Fraud Detection

  As a consequence of the growing swarm of data , fraud has become a major problem leading to great amount of financial loss. Machines are the most profound solution to this problem, for it is impossible for human to analyze hundred of millions of data. Machine learning algorithms aim to find anomalies in the data representing fraudulant activities. Fraud detection problem may be adressed with supervised and unsupervised learning. In the supervised example, the outputs are seperated in to two categories of fraud and non-fraud, and the aim is to predict wether a bank account's activity are indicative of fraudulant activity or not. Semi/Unsupervised learning might be the most feasible method to address this task, due to the high amount of unlabeled data.

• Credit Scoring

  credit score is a numerical value attributed to an individual based on some corresponding features. Credit scores are implemented by lenders, such as banks and credit card companies, to evaluate the risks of lending money. The main goal here is to prevent losses caused from people unable to pay back the borrowed money. Credit scoring models based on statistical analysis are widely used, yet they underperform on big data cases. Therefore, machine learning algorithm are an appropriate alternative that performs well even if the data are complex or large.

• Robo-advisor and portfolio management

  This is an interesting application of ML which involves using algorithms to help individuals plan investements and analyze risks based on their portfolio. By applying these algorithms, people fill in their goals, age, income, and currents financial assets, and the robo-advisor will suggest best area to invest. Automated software agents can use historical results to estimate the best way to allocate your investments.

• Costumer Service

  By now, most people are aware of Amazon's Alexa and Apple's Siri,but digital assistance are also used in finance. Automated phone systems that rely on machine learning can help route callers to the right department within a company, providing good-quality customer service without the need for human employees.

• Stock Price Prediction

  It has always been of interest to find a way to predict future stock value given past observations. Although stock prices are unstable and full of fluctuations, machine learning algorithms can successfully detect complex relationship in past stock prices, and use them to forecast future stock prices.