| Summary: | 碩士 === 國立交通大學 === 理學院應用科技學程 === 106 === The payment of analytical service can discover customer behavior through scientific research. Base on the study, we can describe, explain and establish prediction models, thereby increasing income and reducing expenditure. In the traditional data mining technique, most of them analyze and adjust the variables one by one. It is time and effort consuming, and we can hardly understand the interaction between variables and how it impacts the object. In machine learning technology, the numerical and categorical variable can be computed simultaneously and establish a corresponding model.
The purpose of this research is to compare four machine learning methods and predict customer’s air sampling payments. By comparing different classification models, we aim to determine the best model for the application of AMC analytical service. The result shows that random forest model has the highest accuracy in the testing dataset but time-consuming. At variable importance aspect, the customer11 is the most important variable to classify the labels, followed by country and product part number (PN).
|
|---|
