Risk analysis has become an important AI/machine learning application. At Data Advantage we specialise in risk analytics which includes, fraud detection, credit risk analysis/underwriting, error detection and threat detection.
If you have a need for one of the services below, make an inquiry and we can get put together a free assessment of your situation including a quote for any future work.
Credit risk analytics/ underwriting
We work with some key clients in the financial services industry building their credit scoring/credit risk analysis modelling which helps in their decision-making process around personal loans. Our core advantage in this area is that we find niche features that we have uncovered with experience which seem to persist across various financial products.
The process works by sucking as much information in as possible (bank statements, personal details, social media profile etc) and building a model to estimate the probability of a client defaulting on an underlying product (e.g. a loan)
One of the emerging areas of data science is around fraud detection. This is the process of identifying economic crime. It is estimated that 1 in 3 organisations suffer from economic crime. Common industries where fraud is prevalent historically have been banks, telecommuting and insurance companies, however, these days instances of fraud are much broader than that and are particularly common in e-commerce.
Data science can help in a couple of ways when it comes to fraud detection. First, in pre-emptively identifying areas of risk that might lead to actions to prevent the crime in the first place. Second, data science can help identify areas where fraud is already happening in a timelier manner to help limit the damage that can be done.
We use cutting edge machine learning techniques to reduce instances of fraud. Applications of machine learning vary according the particularly industry or dataset.
Error detection is a niche filed in data science that we have had a great deal of experience in. It’s particularly popular in telecommunications around predicting network outages etc. Such modelling is also becoming popular in energy markets especially around blackout prevention or minimisation.
Threat detection is a form of risk analytics which involves using data science to limit data breaches and attacks which are becoming more sophisticated. Privacy law in many regions are such that the company that gets hacked is usually subject to significant fines, which can be in the millions. Many attacks also come from internal sources. It certainly is a strange world where a company can be a victim of a cyber crime and be liable and often there are no laws or jurisdictions where the perpetrator can be brought to justice. Ok, enough of our rant, lets get back to data science.
The main types of models that are built in this area are probability models using machine learning. For example, if downloads for a day are x then the probability of a threat is y. Machine learning can help is setting thresholds etc.