Challenge to AI x Fintech in Asia

Airitech explores the new technologies and focuses on training AI engineers in Asia. Some of our employees also support as the secretariat of WAI. Feel free to reach out to us by our Airitech Website.
++++++++++++++++++++++++++++++
This article is targeted towards the audience in Japan who are interested in AI but not very good at reading or writing Japanese. I wanted to share the AI journey together with WAI.This article is also available in Japanese and Myanmar Version .
++++++++++++++++++++++++++++++
The third WaiTalk in Japan hosted by Women in AI Japan Team was held on October 8, 2020. I am Khine Zar from Airitech and also a core member of Women in AI(WAI). I would like to share my experience of WaiTalk #3 [ Challenge to AI x Fintech ] Online event. 

Event Name: WaiTalk Japan ( Challenge to AI x Fintech)
Hosted by: WAI Japan Team

About the presenter

Firstly, WAI Japan Ambassador Eriko Toda introduced Ms.Fumiko Inada, the speaker of WaiTalk. She is CEO and co-founder of Bee Informatica Sdn. Bhd, Malaysia. She led a discussion of Challenge to AI x Fintech starting from her biography,　then how she started the Fintech Bee Informatica company in Malaysia and finally shared how her company approached an alternative credit scoring system with AI.
She introduced her company's product ENTREBITION (Digital SME Credit based on Alternative Credit Score for micro-SMEs).

Developing credit scores by ENTREBITION

ENTREBITION targets the owners of SMEs who do not have enough financial and credential records to borrow the loan from normal transitional financial institutions such as banks. So, they develop an alternative credit score which guarantees for those SMEs. Current probability lending ratio of traditional financial institutions to SMEs that run less than 2 years is less than 33.4 % . Those SMEs often take loans from other alternative lenders such as non-traditional local lenders or short-term loan lenders with heavy interest rates to run their business. ENTREBITION is a supportive system for those SME businesses. The main functionality are

1. Digital SME Credit
2. Alternative Credit Scoring

She also introduced DataSoft which is a complete microcredit solution for microcredit organizations. Then she explained the word Credit Scoring Model.

Credit Scoring Model 

Credit scoring model calculates the point or score of the borrower based on his or her credential and personal information. It supports the financial institution to decide if the loan assessment should be or should not be to the dedicated person/SMEs.
The two main decisions of credit scoring models functionality are 

1. Loan to this person/SME is sure to get back
2. How many amount should they decide to loan

Traditional credit score algorithms calculate the score based on company/ business income, salary range and other finance credentials. Even though AI applies to those systems, the ratio for banks' lending to SMEs is still challenging. To solve this problem, ENTREBITION proposes an alternative credit score based on the following 3 models.

1. Psychometric Testing
2. Smartphone Habits
3 Peer Group Scorecard (from same industry colleague database)

Based on their credit score, they guarantee to get the loan for SMEs.
Psychometric Testing[No.1] includes the following 7 personal testing.

1. Comprehension Skills
2. Communication Skills
3. Numerical Ability
4. Motivational Questions
5. Financial Literacy
6. Critical Thinking
7. Relationship

Those are generalized questions and everybody can take the test. This step does not include a data-training process.
SmartPhone Habit[No.2] analyzes the financial responsibilities , capability and plays an important role for creating the credit score model. This step mainly focuses on checking, tracking the borrower mobile’s financial behavior , Utility Bills (Gas, Phone Internet Bill etc) . But mobile habits tracking data are personal data, so they need the owner’s agreement to access those data. Their product had a process for requesting permission from the users to access and track the personal data info.
No.3 is taking from the same industry colleague database data to evaluate the score of current SMEs because it is not enough to generate the complete scoring model just by their data only.
To conclude, No.1 and No.2 are Expert option models and No.3 is Statistical model. Both Expert option models and Statistical models have their own pros and cons. Please refer to the following figure. 

Source: Challenge to AI x Fintech

Their proposed model optimizes the factors from the above mentioned two models and utilizes the hybrid model. As a hybrid model, they have both pros and cons of those two models. She referenced this video from Lenddo EFL to explain those obstacles. Lenddo EFL provides the service non-traditional data comprising social media and smartphone records in order to ascertain customers' financial stability. She mentioned their company also learned a lot from Lenddo EFL.

And then, she also shared her experience of how she started her company Bee Informatica Sdn. Bhd and why they chose Malaysia.
As their application is already released as a beta version, she demonstrated the internal functionality too. You can check their products here. Please note that, currently , Malaysians are their main target and some scoring models are calculated based on Malaysia.
After the presentation, there was a Q/A session, then WAI Japan Team ended the event with their member invitation.

What I learned from WAITalk

This WaiTalk is a little bit different from previous events. From this talk, I get to learn direct field experiences of a founder of an AI based startup. And then, I also learned to use scores from peer databases and apply them to the proposed machine learning model in order to solve the cold start problems in the recommendation/prediction algorithm . Finally, I learned how the hybrid approach in machine learning models can be used in real-world problems.

Next WAI event

I would also like to invite the reader to participate in the next WAI event. For the updated events and news, please follow WAI on Linkedin , Twitter, Facebook and WAI website. To access the WAI Japan Team activity, please follow the WAI Japan Peatix event page and for those who would like to register as a WAI member please follow the following link
https://womeninai1.typeform.com/to/GOcQ3Y
Thank you.