A Study of Using Survey Data to Fuse and Augment the Business Customer Database by Data Mining Technique
| Autor: | Kuo-Wei Chou, 周國維 |
|---|---|
| Rok vydání: | 2011 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 99 Despite the gradual perfection of the establishment of corporate member database, any lack of critical variables might probably hinder the database from yielding the best contribution. The main purpose of this research is to make use of other related survey results or auxiliary database, to help the member database generating more information and to enlarge the value of the member database. This research adopts the integrated data of “Survey on Broadband Usage in Taiwan” between 2007 and 2010, simulates and extracts member database of network companies and utilizes functional mapping to amplify the variable “frequent usage of social networking websites” in the database in order to manifest customer’s Internet usage behavior and increase the revealing information. First, the acquired data are divided into two datasets as internal model establishment and external validity verification by the proportion of 90:10. Then, the established data in the internal model are dispatched into training dataset and testing dataset by the proportion of 80:20. For model construction, all data in the training dataset are to build the full model. Bootstrap is used to perform 100 times of resampling of training dataset, producing 100 single models; three integration models are then developed by mean, maximum and mode. The full model and the three integration models are brought into testing dataset to evaluate the prediction performance of the model in order to indicate the most accepted model. Afterwards, such accepted model is put into the external validity dataset to verify the effectiveness and accuracy of functional mapping by means of confusion matrix. The results demonstrate that functional mapping’s effectiveness of prediction is higher than 70%, suggesting the link of variables is able to accurately and effectively unite the auxiliary data with the target database. Therefore, a corporation only needs to do effective sample survey, builds a prediction model through the investigation of its few members, maps back the observations of the member database, and becomes capable of disclosing more details of the survey. In such a way, not only is the added value of survey being highly raised but also the decision making of a corporation is greatly assisted and the profits of a corporation are hence improved. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|