Predictive Analytics Models Flashcards
6 cards from real ADA practice questions. Tap to flip, then mark Knew It or Still Learning — missed cards come back until you master them.
Read the first 6 Predictive Analytics Models flashcards as text
An audit team wants to develop a model that predicts the monetary value of expected sales for the next quarter based on historical sales data, seasonality, and recent marketing expenditures. Which predictive analytics model would be most suitable for this task?
Answer: Regression Analysis
Regression analysis is the appropriate technique because it is used to model the relationship between a dependent variable (in this case, the monetary value of sales) and one or more independent variables (historical data, seasonality, marketing spend) to predict a continuous numerical outcome.
An internal auditor is tasked with building a predictive model to identify which employee expense reports are most likely to be fraudulent. The model needs to categorize each new report as either 'high-risk' or 'low-risk'. Which of the following models is the most appropriate choice?
Answer: Classification
A classification model is designed to predict a categorical label, such as 'high-risk' or 'low-risk'. It learns from historical data where reports were already labeled as fraudulent or not, and then applies that learning to new, unlabeled data. Linear regression predicts a number, not a category, and time-series is for forecasting trends over time.
Which of the following scenarios is the BEST application of time-series forecasting in an audit context?
Answer: Establishing an expected monthly revenue baseline to identify significant anomalies in the current year.
Time-series forecasting uses historical data points ordered in time to predict future values. In an audit, this is commonly used to create a reliable expectation or baseline (e.g., for revenue or expenses), and then compare actual results against the forecast to flag significant, unexpected deviations that warrant investigation.
An auditor is using logistic regression to develop a model that assesses the likelihood of a company defaulting on a loan. What type of output will this predictive model primarily generate?
Answer: A probability score between 0 and 1 indicating the likelihood of default.
Logistic regression is a specific type of classification algorithm used to predict a binary outcome (e.g., default/no default). It works by calculating the probability of the event occurring, which is expressed as a value between 0 and 1.
When an auditor develops a supervised machine learning model to predict fraudulent transactions, what is the most critical requirement for the training data?
Answer: It must contain a large and representative set of both known fraudulent and non-fraudulent transactions.
Supervised learning models require a labeled dataset to learn from. To effectively learn the patterns that distinguish fraudulent from legitimate transactions, the model must be trained on a comprehensive dataset that includes clear examples of both categories.
An auditor uses a predictive model to flag purchase orders that have a high probability of being unauthorized. The model identifies 100 high-risk POs. Upon manual review, the auditor finds that 15 of these were indeed unauthorized. In the context of evaluating the model's performance, what does the number '15' represent?
Answer: True Positives
True Positives are the outcomes where the model correctly predicts the positive class. In this scenario, the 'positive' class is an unauthorized PO. Since the model correctly flagged 15 POs that were confirmed to be unauthorized, they are True Positives.