Frequently Asked Questions
This section provides answers to common queries regarding the TimeDetect API.
What kind of anomalies can TimeDetect find?
TimeDetect is a robust and customizable API that can find anomalous patterns in single registrations, the total for an employee for a given day and the absence of registrations for an employee for a given day. The fields considered for anomaly detection are not limited to the common fields, such as start time, end time and work duration, but can also include custom fields, such as overtime, sub-project, machine type and driving distance. The custom fields are specified in the list of numericals.
How does TimeDetect distinguish between different types of employees?
TimeDetect builds many specialised models for different departments, work types, projects and employees and uses the most specific one when looking for anomalies. This means that a part time employee will get different results than a full time employee for the same registration.
What is the difference between the create prediction endpoint and the real time prediction endpoint?
The create prediction endpoint is designed to make predictions on a large batch of registrations, e.g. 4,000 registrations, including aggregated results. The real time prediction endpoint is designed to give predictions for a small number of registrations fast and does not include aggregated results. A typical use case for this is when an employee is about to or just submitted a registration and would like real time feedback.
Why am I not receiving missing scores for some of my employees?
Missing registrations is exclusively handled by the aggregated model. If you're getting predictions but not the anticipated missing scores, it might be because the employee does not have an aggregated model. It is also worth noting that we do not generate aggregated results for registrations in the future. To understand the requirements for obtaining an aggregated model, please see the model section.