Synechron experts on three data science trends to watch

23 June 2021 Consultancy.eu 3 min. read
Profile
More news on

In the latest edition of its annual ‘Top Strategic Technology Trends’ report, Synechron heralds artificial intelligence and data science as one of its eight major trends for 2021. Experts from the firm’s Dutch practice outline three of the data science developments fuelling this trend. 

Hachem Ohlale, Head of Analytics & Data Science

“A long standing topic in artificial intelligence (AI) is how to reduce the reliance of AI models on large quantities of classified data,” said Hachem Ohlale, Head of Analytics & Data Science at Synechron.

“One solution we see a trend towards is the use of Generative Adversarial Networks (GANs), which can generate large amounts of realistic synthetic data from small samples. GANs have been most infamously used for generating fake pictures or artwork, but are increasingly being used in generating financial time series data, like stock prices and customer data, etc.”

Hachem Ohlale, Cameron Peak, Arjen Koomans

He added, “Feeding your models with a mix of real data, and this high-quality synthetic data, can enable more problems to be approached with AI, and improve accuracy and robustness in your existing models.” 

Ohlale continued, “Another area of interest is more business applications of self-supervised models, like those behind Alpha Go Zero, with AI teaching itself to solve a problem without human classification of the data. If you can avoid manually classifying thousands of data points, you can achieve results faster, and spend more time realizing results in your data.”

Cameron Peak, Lead Data Scientist

“In recent years, we have seen many businesses realise the potential of unstructured textual data to transform their businesses through automating the handling of vast data, such as for legal documents, corporate filings, email alerts, client correspondence and news sources,” said Cameron Peak, Lead Data Scientist at Synechron.

“Each year has brought generational shifts in the possibilities of Natural Language Processing (NLP) techniques, for example GPT-3 and BERT, with language models that can now match or exceed human performance in comprehending and classifying text. This year I expect to see this trend accelerate, with wider adoption of Natural Language Generation (NLG), which uses AI to create many of the hand-produced documents we interact with every day.” 

“By deploying the summarisation capabilities of NLG, you can deliver highly personalised reports and analysis to users in an easily digestible form, as well as automating the repetitive and time consuming production of reporting and workflow documents,” he added.

Arjen Koomans, Director Client Lifecycle Management and Financial Crime

Data Science tools will strut their stuff in 2021, particularly as they relate to Know Your Customer (KYC) and financial crime purposes and compliance, predicts Arjen Koomans, Director Client Lifecycle Management and Financial Crime at Synechron. 

Technologies like Machine Learning (ML), Optical Character Recognition (OCR) and Natural Language Processing (NLP) will prove to be highly useful by increasing efficiency, reducing costs and detecting financial crimes. “Building a KYC file requires gathering of data from various sources – such as data providers, client documents and public sources -- but many of these sources are still paper based or unstructured.” 

“What’s more, while processing of structured data is quite standard today, processing unstructured data, such as from annual reports, is still very challenging. But tools like OCR and NLP can efficiently extract insight from unstructured data,” Koomans explained. 

In addition, the key to detecting financial crime is analysing client behaviour, which has traditionally been rule-based and is rigid in monitoring and in finding the ‘unknown unknowns’, he added. “Machine Learning is the way forward for advancements in financial crime detection as it supplements or even supersedes existing rules-based detection systems by developing specific algorithms trained to detect specific types of financial crimes,” Koomans noted.