Big data and machine learning improve the efficiency of the team. How do they do it.

[CUK77]

Why are large resources invested in the transformation of the collection?

Collection is definitely not the most attractive part of the credit system. And not the most pleasant one for both sides of the process. But it is absolutely necessary - the credit system cannot exist without it. That is why now large resources are being invested in its transformation. The goal is to make the process less stressful for both parties and increase its efficiency.

Both banks, MFOs, collectors and startups are working on solving the problem. TrueAccord (USA), Attunely (USA), Dasceq (USA), Symend (Canada), CollectAI (Germany), inDebted (Australia), Ziyitong (China), Flow (Singapore) and others are transforming the collection through innovative approaches. And at the heart of all of these approaches are big data and machine learning.

Having studied in more detail what exactly these startups are doing, I have identified 3 main areas for the application of big data and machine learning in the collection industry.

Additional Information.
The Chinese company Ziyitong has managed to recover about $ 29 billion in debt since it opened in 2016. Its clients include about 600 debt collection agencies and more than 200 creditors. The company is based on a machine learning system that includes two main elements.

The first is the collection of data about the borrower and his environment from open sources (mainly social networks). The system analyzes the information found, and then uses it in telephone contacts with the borrower and his friends.

The second is the personalization of communication with the borrower. The system records the conversations, analyzes them, and then determines the wording that is most likely to compel the borrower to repay the debt.

The return rate of such a system is 41%, which is on average 2 times higher than that of traditional collection methods in China (20%).

Personalization of time, return conditions, frequency and communication channel
This is the most popular area, on the development of which all of the above startups are working.

So, TrueAccord is trying to soften the collection process for borrowers as much as possible. To do this, the company has completely abandoned phone calls and sends emails to borrowers with personal return conditions, and reminds them about them using text messages and targeted Facebook ads.

CollectAI, a German startup, is also looking at message sending times, email content and opening conversions, which helps it create dynamic landing pages for each individual borrower. And these actions are yielding results - within the framework of a 6-month cooperation with Hanseatic Bank, the bank's overdue debt recovery rate increased by 24%.

Personalization of messages.
From the whole area of personalization of communication with the borrower, it is worth highlighting the personalization of messages, which is used by most of the above startups. It is technically the most difficult element, but at the same time it is the most effective. To implement it, you must complete the following steps:

• start recording all conversations with borrowers, track their effectiveness;
• use an audio transcription system and with its help translate telephone conversations into text;
• compare 3 types of data with each other: the type of the borrower, the text of telephone conversations, efficiency;
• to build the type of the borrower, collect external data (first of all, information from social networks);
• on the basis of 3 types of data, train an algorithm that predicts the optimal messages for each type of borrower.

Such a system will make communication between the collector and the borrower less stressful for each of them, as well as increase the level of repayment of overdue debt.

 

[Lord777]

Smart antifraud: How Big Data and Machine Learning Protect Your Money​

In continuation of the topic about preventing and solving crimes using IT, today we will tell you what antifraud systems are, why they are needed and where they are used. We will also look at the role of Big Data and Machine Learning technologies in such fraud detection tools. Read our article on why big data and machine learning automate the monitoring and detection of financial and accounting violations, preventing money theft and other transactional crimes.

What is antifraud and where is it used?
Usually, anti-fraud (from the English anti-fraud) is associated only with the banking sector, when financial transactions are evaluated for fraud, for example, when a payment card is used by an attacker, and not by its owner. However, it is not just credit institutions and online stores that need antifraud. An interesting example is that of a large gas station chain, where a profitable loyalty program was launched with a partial refund for purchases, the so-called cashback. At one of the gas stations, the operator made all payments through his personal card with such cashback, accepting cash from customers and sending it through his account. Although this case nominally does not violate the law, in fact it is insider behavior of staff in their own interests. The deceptive scheme was discovered manually: after analyzing all transactions for the day, it was found that almost a tank of gasoline was purchased with the same bank card. The anti-fraud system would identify and freeze such transactions automatically.

Another classic case is the field of insurance, in particular, when an insurance agent takes several policies and does not register them in the accounting system. At the same time, he tells the client about a big discount and sells the policy cheaper. And when the client has an insured event, the agent will register his policy retroactively. At the same time, other similar policies will remain unregistered, as a result of which the agent will receive a huge benefit. Thus, anti-fraud is relevant not only for banks, but also for other businesses where online commodity exchange relations arise and money transactions occur.

However, in practice, such anti-fraud systems are most in demand in banks and online stores, as this is where most online payments are made. In particular, in Sberbank alone in 2018, an anti-fraud system based on Big Data analytics helped save more than 32 billion rubles belonging to depositors. The global market for anti-fraud systems is expected to exceed $50 billion in 2024. Today, the following companies are considered the leading suppliers of antifraud systems:

IBM (USA); FICO (USA); SAS Institute (USA); BAE Systems (UK); NICE Systems (Israel); LexisNexis Risk solutions (USA).

How does the Fraud Detection system work and what does machine learning have to do with it
The anti-fraud service includes standard and unique rules, filters, and lists for checking each transaction. In particular, the following restrictions are most popular:

· the number of purchases made with one bank card over a certain period of time;

* maximum amount of a one-time purchase for one card in a certain period of time;

· number of cards used by a single user in a given time period;

· number of users using a single card;

* accounting of the history of purchases by bank cards and users ("black" or" white " lists).

The following filters are also often used:

* validators, for example, checking bank card details for correctness;

* geography, when the IP address from which the user is trying to make a purchase is associated with a specific country. For example, some African countries have a high level of skimming and card compromise, so payments from these countries are highly likely to be fraudulent.

* matching parameters, such as the country of the payer's IP address and the bank card issuer. If the cardholder did not inform the bank in advance about their travel, then in case of making payments from another country, there is a high probability that the banking details were stolen and used by intruders.

* stop lists when the card has already been detected in fraudulent transactions or its owner has reported data compromise to the issuing bank.

To implement such standard rules, fast user recognition is performed using various parameters and algorithms, including using machine learning. And thanks to Big Data technologies, a data set is formed for automatic intelligent assessment of consumer behavior. Information about the transaction itself and its metadata (sender and recipient of the payment, amount, time, place, additional information) is aggregated and compared with the history of previous payments. As a rule, true information about the cardholder and a set of user parameters correspond to standard patterns of behavior of decent buyers. These additional factors are also taken into account by anti-fraud services when calculating the probability of fraud.

This is how Machine Learning tools form a template for user behavior, using clustering algorithms to determine the most typical amount of cash withdrawals or purchases for this client. Unlike the filters and constraints discussed above, self-learning ML models are able to extend previously defined rules, adapting to the client. However, this flexibility does not conflict with precision. If an individual transaction does not fit into the previously formed templates, taking into account possible assumptions, it is considered an anomaly. For example, a person suddenly withdraws or spends an uncharacteristic amount. Another suspicious case is when several payments to different locations with the same amount leave the same account at once. Another sign of fraud is the transfer of small amounts to many different accounts.

Thus, machine learning algorithms allow the anti-fraud system to conduct continuous monitoring and detection of suspicious cases, providing flexible configuration of filtering parameters through interactive formation of behavioral patterns. In addition, Machine Learning tools automate decision-making by rejecting abnormal operations and blocking compromised cards.

In the next article, we will continue to look at anti-fraud systems and explain how the tools of graph analysis of big data embedded in them help to uncover criminal money laundering schemes.