首頁 > 網路資源 > 大同大學數位論文系統

Title page for etd-0831110-045612


URN etd-0831110-045612 Statistics This thesis had been viewed 1890 times. Download 12 times.
Author CHUNG-HSIEN CHANG
Author's Email Address ssc111h@hotmail.com
Department Information Management
Year 2009 Semester 2
Degree Master Type of Document Master's Thesis
Language Chinese&English Page Count 114
Title Applying Data Mining to the Sales Forecasting of
Smart Phone in Distribution Industry
Keyword
  • Distribution Industry
  • Smart Mobile Phone
  • Sales Forecast
  • Data Mining
  • Back Propagation Neural Network
  • Back Propagation Neural Network
  • Data Mining
  • Sales Forecast
  • Smart Mobile Phone
  • Distribution Industry
  • Abstract Abstract
      Mobile phone development in recent years is really amazing, its keeping evolution has gradually changed our lives and becoming an indispensable tool for a modern technological savvy human being. The mobile phone market is highly competitive and the product is seen as: high unit price, short life cycle, intense price competition with fast diminishing value as new models are introduced, high product substitution, little consumer loyalty, especially seen in the high end smart phones. A smooth logistic channel between the retailers and manufacturers is the key to maintain a just in time inventory level without large overhead and warehouse space. Too much inventory is detrimental to the company cash flow and too little inventory is lost opportunities when consumers can’t get the product right away and therefore, an accurate forecast of sales for enterprise operations become more important.
      In this study, high-end smart mobile phones sold between 2006 and2007 are selected as research objects. The historical sales data for high-end smart mobile phone features, such as historical sales volume, brand, model, price, time, area, and the distribution rate of economic stimulus measures indicator, the consumer confidence index, stock-weighted index and the dollar exchange rate and other economic factors that may affect the prediction model are considered as variables. Through various forms of feature selection, the construction of a variety of back-propagation neural network model (BPN), and then a minimum mean absolute percentage error (MAPE) as the evaluation criteria, the best sales forecasting model is obtained.
     
      The results showed that the present study and the accuracy level is very close to the actual sales, which can help build the optimal sales forecast model of high-end smart mobile phone for distribution industries , and can be further embedded into an enterprise knowledge management to accumulate experience, and thus enhance business performance and improve their competitiveness.
    Advisor Committee
  • Yen-Ju Yang - advisor
  • Files indicate in-campus access only
    Date of Defense 2009-01-25 Date of Submission 2010-08-31


    Browse | Search All Available ETDs