Log in/Register Request a Demo Instruction Manual
A new AI-powered service that helps businesses or financial institutions or fraud decision experts make binary and multi-class decisions identifying fraudulent activities, transactions, or applications by performing binary or multi-class predictions and classifications for example, or sanity checking their existing assessments, binary or multi-class predictions and classifications. Some examples:
1. Detecting accounting/tax fraud: Butterfly AI can help auditors and relevant authorities to detect accounting or tax frauds by detecting anomaly patterns within data. Knowing the key data features of hundreds of past audited accounts including total current assets, trade account payables, total assets, total common/ordinary equity, cash and short-term investments, cost of goods sold, common shares outstanding, total debt in current liabilities, long-term debt issuance, total long-term debt, depreciation and amortization, income before extraordinary items, total inventories, other investment and advances, total short-term investments, total current liabilities, total liabilities, net income (loss), total property, total plant and equipment, total preferred/preference stock (capital), retained earnings, total receivables, sales/turnover (net), sale of common and preferred stock, income taxes payable, total income taxes, total interest and related expense, fiscal annual price close, WC accruals, RSST accruals, change in receivables, change in inventory, percentage of soft assets, depreciation index, change in cash sales, change in cash margin, change in return on assets, change in free cash flows, retained earnings over total assets, earnings before interest and taxes over total assets, actual issuance and book-to-market and any other potentially relevant data features and having verified, confirmed and labelled each one of those past audited account cases as (0: Not Fraudulent and 1: Fraudulent) i.e. labels and targets of classification and detection process, decide whether a new audited account is fraudulent or not (and also present the probability and certainty of fraud) provided that the data features (described above) are available to the classification and detection process.
2. Fraud detection at self-checkouts in retail and supermarkets: To enable retailers and supermarkets identify cases of fraud within the self-checkout process and minimise revenue loss and disruption to innocent customers. Butterfly AI can help follow-up checks to detect the fraudulent activity seamlessly. Knowing data features of many past self-checkout shopping journeys including customer’s historic individual trust levels. 6: Highest trustworthiness, 0: Lowest level of trustworthiness (through credit and shopping history, etc.), total time in seconds between the first and last product scanned, grand total of products scanned, number of voided scans, number of attempts to activate the scanner without actually scanning anything, number of modified quantities for one of the scanned products, average number of scanned products per second, average total value of scanned products per second, average number of item voids per total number of all scanned and not cancelled products and any other potential relevant data features including weight, and having all those past journeys correctly labelled as fraudulent (1) or not fraudulent (0), decide/classify whether a new self-checkout journey (knowing all the above features for current customer) is fraudulent or not (and with what probability).
3. Mobile payment fraud detection: Having recorded the details of thousands of past mobile payments including the payment type (cash-in, cash-out, debit, payment and transfer), amount of the transaction in local currency, ID of customer who started the transaction, the time within which customer started the transaction, the location of the customer who started the transaction, the type of mobile contract of that customer, initial balance before the transaction at the origin, customer's balance after the transaction at the origin, the number of transitions that the customer has had during past year, ID of the customer who received the transaction, the location of the customer who received the transaction, the type of mobile contract of the receiving customer, initial balance before the transaction at the destination, customer's balance after the transaction at the destination, the number of transitions that the receiving customer has had during past year, and having verified, classified and labelled each one of those past mobile transactions as fraudulent or not fraudulent, decide whether a new mobile transaction is fraudulent or not (and with probability/certainty) provided that the data features (described above) are available for that new transaction.
4. Auto insurance fraud detection: Knowing the data feature of past car insurance claims including months as customer, age, policy number, policy bind date, policy state, policy csl, policy deductible, policy's annual premium, umbrella limit, zip, insured sex, education level, occupation, hobbies, relationship, capital gains, capital loss, incident date, incident type, collision type, incident severity, authorities contacted, incident state, incident city, incident location, hour of the day within which incident happened, number of vehicles involved, property damage, bodily injuries, witnesses, police report available, total claim amount, injury claim, property claim, vehicle claim, auto make, auto model, auto year and knowing each one of those past claims has been fraudulent or not, decide whether a new auto insurance claim is also fraudulent or not and with what probability.
5. Credit card fraud transaction detection: Knowing the data features of past credit card transactions including the distance from home where the transaction happened, the distance from last transaction location, ratio of purchased price transaction to median purchase price, whether the transaction happened from same retailer or not, whether the transaction went through chip (credit card) or not, whether the transaction happened by using PIN number or not, is the transaction an online order and knowing whether each one of those past credit card transactions was fraudulent or not, decide whether any new credit card transaction is also fraudulent or genuine and with what probability.
6. Banknote authentication: Imagine the key features of digital images of banknotes have been extracted into a structured csv data file. Knowing that those data features belong to some genuine and forged banknotes, including Machine Identifier, variance of image, skewness, kurtosis, entropy, colours, etc. and knowing whether each one of those banknotes are forged or genuine (ML labels), decide whether any new banknote is also forged or genuine.
and any other binary or multi-class fraud-related classification assessment or prediction use cases.
Log in/Register Request a Demo Instruction Manual
