Guide
1. Dataset
1.1 Training Data
Training data is a combination of all employee data samples with a non-null attrition attribute and all employee data history. The training data has two target classes: “no” and "yes". The details are as follows:
| Class (Attrition Value) | Description |
|---|---|
| no | employee data samples with non-null attrition attribute values + employee history data samples whose attrition attribute value is “no”. |
| yes | employee history data samples whose the attrition attribute value is “yes”. |
1.2 Test Data
Test data are samples of employee data where the attribute value is null. This is explained in more detail in Section 3.1.
2. Login
Use this identity to log in to the system.
- Username: fwaskito
- Password: fwaskito123
3. Employee Data Page
3.1 Edit Data (Provide Test Data)
A test data sample can be created by editing one or more attribute values of an
existing employee data sample. This is because, in principle, test data is different (not
identical) from any training data.
An employee data sample that has its value changed will automatically have its attrition
attribute value set to null. Does this mean that the amount of training data will decrease?
It doesn't. This is because each data sample is stored as employee data history data when a
change action (update or delete) is performed on it.
3.2 Delete Data
A deleted employee data sample will automatically have its attrition attribute value changed from "no" to "yes". After that, the data disappears from the employee data and becomes the new employee history data.
3.3 Add Data
When a new employee data sample is added, its attribute value is set to "no" by
default.
Why doesn't this new data become new test data? Its attribute values would make it different
from the samples in the training data, right? Yes, it would. But the current functionality
of the system focuses on predicting attrition only for data that is assumed to be available
and integrated with the company's HR database, not for random dummy data.
4. Employee History Data Page
Employee history data is employee data that has had an action (update or delete) performed on it. Any employee history data resulting from an update/edit action can be replaced by employee history data, with the same employee ID, resulting from a more recent update action. Therefore, one employee data can generate a minimum of 1 (update) and a maximum of 2 (update and delete) employee histories.
5. Prediction Data Page
5.1 Make A Prediction
There are two options for prediction: predict and predict all. Predict is single data prediction for one of test data. While predict all is a prediction for all available test data.
5.2 Reset Prediction Data
Reset prediction data is an action that sets the value of the attrition attribute to
"no" for all test data. As a result, this action will clear all available test data.
Note: the prediction action is a temporary action, where the results are only stored while the user is logged in and deleted after the user logs out. To save the prediction results permanently, we can export the prediction data to a document in the available format.