METHODOLOGY PROSIDING 2nd ACISE 2015

Semarang, 7 Oktober 2015 319 No Attribute Attribute of Customer Service Weight Value 16 Quick and appropriate in complaint response 0,0472 17 Staff provide information about new products -0,0044 18 Discounts for loyal customers 0,0231 Attribute with the largest weight value In Table 1 shows that the attribute 3, 5, 6, 10, 11, 12, 14, 15, and 17 have negative weights value or is below the specified threshold. This indicates from 20 respondents, these attributes are irrelevant on the level of customer satisfaction with the services. Attributes that affect customer satisfaction levels are ordered beginning with largest value. The attributes are 16, 4, 18, 9, 7, 1, 13, 2, and 8. These attributes will be used as a dataset for DT calculation. Result of Service Level Analysis The dataset used in the analysis of DT are dataset which the attributes affect to the level of customer satisfaction based on the results of Relief analysis. The DT calculations from the dataset show that attribute 4 which is “No damaged product when it arrives to customer” became the root attribute. In performing splits, DT calculated value in attribute 4 which do not have a homogenous class. The next calculation generated attribute 7 “Competent staff and good distribution facilities” became internal node of value 4 Satisfy at attribute 4. DT calculation stopped because of the values that belongs attribute 7 could not be split. This is due to the value of the next gain result does not generate value with a significant difference between one attribute to another, so the split stopped. Fig 1 shows the structure of DT. Figure 1. DT structure result Pattern of customer satisfaction level will be established based on structure shown in Figure 4. The pattern of customer satisfaction divided into two groups, where there are a pattern of customers at a rate “Quite satisfy” with the service and “Satisfy” with the service. Patterns of customers with the “Quite satisfy” level described in Table 2 and the pattern with a “Satisfy” level in Table 3. Table 2. Pattern of Customer with “Quite Satisfy” Level No Rule “Quite satisfy” 1 IF Attribute 4 = 2 Less satisfy THEN Overall customer satisfaction is “Quite satisfy” 2 IF Attribute 4 = 3 Quite satisfy THEN Overall customer satisfaction is “Quite satisfy” 3 IF Attribute 4 = 4 Satisfy AND Attribute 7 = 2 Quite satisfy THEN Overall customer satisfaction is “Quite satisfy” Table 3. Pattern of Customer with “Satisfy” Level No Rule “Satisfy” 1 IF Attribute 4 = 4 Satisfy AND Attribute 7 = 3 Quite satisfy THEN Overall customer satisfaction is “Satisfy” 2 IF Attribute 4 = 4 Satisfy AND Attribute 7 = 4 Satisfy THEN Overall customer Semarang, 7 Oktober 2015 320 No Rule “Satisfy” satisfaction is “Satisfy” 3 IF Attribute 4 = 4 Satisfy AND Attribute 7 = 5 Very satisfy THEN Overall customer satisfaction is “Satisfy” Tables 2 and 3 describe the pattern of customer satisfaction level to existing services today. For example, the first patternμ If customer feel “Less satisfy” about performance of attribute 4, then customer satisfaction level is “Quite satisfy” to the overall performance of services. Depictions like these examples apply to each pattern in the table. The advantage of the research is provide information to company about the services that impact on customer satisfaction and the pattern of customer level satisfaction. The disadvantage of this research is only can be applied in specific time. Customer satisfaction toward to the service always change along with time, and it’s depend on quality of service that company provide. So the result of research will be different when performend in different times.

4. CONCLUSION

There are nine attributes affect the level of customer satisfaction identified by using Relief. From the nine attributes, DT analysis generates six patterns of customer satisfaction with the services provided.

5. RECOMMENDATION

Application in web or phone that contain the analysis and connect between company and customer required to be develop to perform analysis in anytime and anywhere. REFERENCES Berry M, Linoff G. 2004. Data Mining Techniques: For Marketing, Sales, and Customer Relationship Management, 2 nd edition. Canada CA: Wiley Publishing Bramer M. 2007. Principles of Data Mining, 1 st edition. London UK: Springer. Han J, Kamber M, Pei J. 2012. Data Mining Concepts and Techniques, 3rd edition. Waltham US: Morgan Kaufmann. Karten N. 2003. How to Establish Service Level Agreements. [downloaded on] www.nkarten.com . Accessed on 05 May 2015. Kira K, Rendell L. 1992. The feature selection problem: traditional methods and a new algorithm., AAAI- 92 Proceedings. Kononenko I. 1994. Estimating Attributes: Analysis and Extensions of RELIEF. Journal Machine Learning ECML, vol 784. Kononenko I, Sikonja MR. 1997. An Adaption of Relief for Attribute Estimation In Regresison. Journal Machine Learning NSW Premiers Department Corporate Services Reform Team. 1999. Service level agreements guidelines for public sector organisations. NSW US: City Design and Production. Purnasari H. 2012. Evaluasi Layanan Transaksi Pelanggan Berbasis Quality Function Deployment. [Undergraduate Thesis]. Bogor ID: Bogor Agricultural University. Rygielski C, Wang JC, Yen DC. 2002. Data Mining Techniques for Customer Relationship Management. Journal Technology in Society. 24: 483 –502. Tsiptsis K, Chorianopoulos A. 2009. Data Mining Techniques in CRM: Inside Customer Segmentation. West Sussex GB: Wiley Publishing.