RBC files ‘purchase prediction’ patent as tech firms encroach on banking turf
TORONTO The Royal Bank of Canada is looking to patent several artificial-intelligence inventions, including one that could predict when consumers are about to make major purchases, public filings show.
Toronto-based RBC, Canada’s biggest bank, has filed several Ai-related applications that were made public in December, according to the Canadian Intellectual Property Office’s patent database.
One patent application describes a computer system that could receive transaction data and create “structured” data in which purchases are identified and labelled. A “recurrent neural network” would then be trained to build a model that could be used to predict the likelihood of a purchase.
A deeper description of the system in the filing — which notes the scope of the application is not meant to be limited to its various “embodiments,” the specific examples of how the invention could be used — shows a computer processor receiving raw data from sources such as retailers or banks. As another example, “bureau information” about loans that were taken out by clients could be fed in as well.
“Labelling transactions, for example, when a client has or has not purchased a car based on transaction history ... may be used for training to allow for deep learning models to learn from example,” the filing states.
There could also be labelling done through “data mining,” the filing suggests, helping to identify a product purchased based on who the merchant is and the amount spent.
“By way of example, it may be determined that an individual who spends $40,000 at a car dealership has likely purchased a vehicle, and not branded car dealership merchandise,” the description says.
Once a model is trained and stored it “may be used to provide inference for purchase prediction, for example, on new clients who are waiting to be scored on their likelihood to buy a vehicle, a home, a vacation, or the like, and thus predicts the likelihood of various upcoming transactions,” the description says.
A similar setup could also be used to help direct “targeted and relevant” advertising campaigns at clients, “who would otherwise find campaigns an annoyance,” the filing states.
The patent application, for a “K-LSTM (long short-term memory) architecture for purchase prediction,” was made public in mid-december, according to the database.
RBC is no stranger to artificial intelligence, and the lender has its own research institute, Borealis AI. Other banks are invested in the technology as well, such as Toronto-dominion Bank, which bought Canadian AI company Layer 6 Inc. in 2018.
Royal Bank also has experience in “predictive technology.” In 2017 RBC launched two “NOMI” digital services that a press release noted “use a client’s account activity to identify trends, unusual activity and potential savings opportunities.”
RBC’S chief executive, however, has voiced concern over big technology companies finding ways to get between the lender and its customers. The bank used to be among the first to know about the new financial needs of their consumers, RBC CEO Dave Mckay noted to shareholders in 2018, but now those clients are increasingly announcing their plans to big tech firms via internet searches and social media.
“Moving higher up in the funnel of decision-making and life moments before the financing decision or before the banking decision is mission-critical in us taking control in that journey and not being beholden to somebody else’s platform,” Mckay said in March 2019, according to Bloomberg.
RBC’S latest patent applications would appear to fit into that mission. Patents provide the exclusive right to make, use and sell new or improved products, machines and processes.
Other patent filings made by RBC with the same dates as that of the purchase-predictor include those titled “correcting bias in supervised machine learning data,” “portfolio-based text analytics tool” and “system and method for processing natural language statements.”
The portfolio-based application gives an example of an investment adviser trying to “keep up” with global currencies. The adviser could conduct a search, with a platform pulling together articles and processing them to identify certain themes.
“For example, this may have enabled prediction of themes such as ‘cryptocurrencies,’ which can impact investment decisions going forward,” the filing says.