ANALISIS DAN PREDIKSI KLAIM GARANSI SEPEDA MOTOR DENGAN MELIBATKAN DATA FOLLOW-UP

• Main Author: ANANTASARI
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/1905
• Institution: Institut Teknologi Bandung

A warranty is a contractual agreement, which requires the manufacturer to rectify all item failures that occur within the warranty <br /> period. As a result, by selling a product with warranty means additional cost due to servicing of the warranty. This warranty cost is one of the <br /> components of product price. If the warranty cost estimation is higher than the actual cost, the product price becomes high and uncompetitive. <br /> On the other hand, if it is lower then it will reduce the company.profit. Because of the company has to cover the higher warranty claim. As a result, it is important for the manufacturer to estimate warranty cost <br /> accurately. In this thesis, we will estimate the warranty cost for a motorcycle produced by PT X. The warranty policy offered is a two-dimensional <br /> warranty, which one dimension represents age and the other usage. In addition to this, to help the company to make a budget for warranty <br /> claims, it is also important for the manufacturer to predict the future warranty claim. Therefore, in this thesis we will also predict the future warranty claims for the motorcycle. The important step in estimating the warranty cost is to model failures of the motorcycle. There two approaches for modelling failures. The first approach is two-dimensional approach which considers failures as points process in a two-dimensional plane, but this approach involves a higher complexity. The second is one-dimensional approach which <br /> views usage as a function of time and modelled failures as a onedimensional point process. In this thesis, we follow the second approach <br /> and view the motorcycle as a system. Parameter estimation for the failure (or reliability) model utilized claim and follow-up data and this results in a more representative reliability model compared to if we only used the claim data. The resulting parameters are then used to estimate the warranty cost as well as to predict the warranty claims. In this research, we carried out analysis of claim and follow-up data <br /> for the motorcycle sold in 2001 in Bandung city and claimed during 2001. The results of this research show that the motorcycle failure distribution and warranty cost distribution have a positively skewness and hence the result with usage censoring approach II and median approach give a better warranty cost estimation. For prediction of future warranty claims, the use of the usage censoring approach II gives also a more <br /> representative prediction.