Skip to main content

Outlier Detection - Deposits - Others - Dollar Value

Description

This test identifies anomalous transactions by comparing recent deposit values against other users' historical patterns. It calculates symbol-specific statistics, including mean and standard deviation of transaction dollar values, and flags outliers that significantly exceed certain thresholds. Transactions are analyzed in the context of the aggregate behavior of others for a given symbol.

Use Cases

  • Fraud Detection: Identify users engaging in potentially fraudulent activities, such as sudden and uncharacteristically large deposits, which deviate significantly from their historical transaction behavior or the average activity for a specific symbol.

  • Risk Management: Flag anomalous transactions that may indicate operational risks, such as incorrect transaction processing, system errors, or misuse of accounts, enabling financial institutions to mitigate risks proactively.

  • Anti-Money Laundering (AML) Monitoring: Detect unusually large transactions that could indicate money laundering activities.

Required Data

Deposits Data

VariableDescription
timestampTimestamp of the deposit made (YYYY-MM-DD hh:mm:ss).
user_idUser ID to identify the individual user.
currency_typeThe currency type of deposit being made; fiat or crypto.
symbolThe asset symbol; e.g., BTC, ETH, USD, EUR.
price_usdThe price of the symbol in USD.
amountThe amount of the symbol being deposited.

Parameters

ParameterDescriptionTypeDefault ValueConfigurable
analysis_windowOverall duration (days) for analysis.Integer1 (days)Yes
historical_windowNumber of days to consider for historical data.Integer90 (days)Yes
historical_minimum_number_transactionsMinimum number of historical transactions per user required.Integer5Yes
analysis_minimum_aggregate_dollar_thresholdMinimum aggregate dollar value of transactions per user in analysis period.Float500Yes
historical_minimum_number_daysMinimum number of unique active days required.Integer2Yes

Methodology

  • Compute Historical Statistics For each user and symbol, calculate the historical mean and standard deviation of transaction dollar values. Additionally, compute aggregate symbol-wide statistics, including the mean of historical averages and the median of standard deviations for each symbol.

  • Symbol-Wide Outlier Detection Compare a user’s transactions against the historical behavior of others for the same symbol. Transactions are flagged if their value deviates significantly from the symbol-wide mean plus a defined threshold.

  • Categorization and Reporting Flagged transactions are classified as symbol-wide anomalies. These are then prepared for reporting, either as CSV exports or tickets for further investigation.