CRISP-DM: Modeling and Evaluation Analysis Essay Sample

Milestone 3
DAT 690
Student Names
CRISP-DM: Modeling and Evaluation
Introduction
The milestone describes the selected model and the expected results. Next, the model will be run, and the captured results will be documented. An iterative modeling process will be used to derive results that support the final model design. Lastly, charts and statistical model outputs will be captured and documented.
Project Overview
The Cross-Industry Standard Process for Data Mining (CRISP-DM) process will be followed throughout the data preparation and transformation process. The CRISP-DM phases of data description, preparation, cleaning and integration, and reformatting. Data was summarized used descriptive statistics and correlation analysis using R Studio. Descriptive statistics and correlation analysis helped to collect, collate and analyze the GE employee dataset.
GE Dataset Evaluation Process
The dataset had to be cleaned and reformatted for ease of use in R studio. The process of preparing the employee dataset adhered to all the CRISP-DM iterative processes. The external dataset used was prone to errors that adversely affected its quality, accuracy, and reliability. Firstly, data modeling will be achieved by changing the categories into numeric values for ease of input into the R studio (Mohd Selamat et al., 2018). This process entailed using the transform command in Rattle, while the second step in data cleaning involved riding the dataset of unnecessary details. The variables obtained were hires, layoffs_discharges, other_separations, and quit. These input variables can accept numeric input values. Based on these variables, linear modeling was used to determine attrition levels for GE employees. Nevertheless, some variables such as Hires were found to have no significant impact on the prediction of employee attrition.
The data cleaning phase was used to clean the data variables in the dataset iteratively. Changes made to the data element column involved altering input values for Quit [1] and All other [0]. Besides, data cleaning processes entailed changing other categories such as other_separation and Layoff_Discharges. Next, linear modeling was used where actual data was taken from the GE employee dataset and used as input in the linear data model. Subsequently, that eliminated inadvertent gaps, thereby increasing the accuracy and reliability of the model (Exenberger & Bucko, 2020).
Figure 1: Linear Model on Attrition [Validate/Proportions]
10763253781425
Based on the GE dataset, out of 164 hires, there were a total of 133 total separations that includes layoffs (24), other separations (77), and those that Quit (48). The error matrix for the linear model on attrition was derived for variables: Hires, Layoff...