ARTIFICIAL NEURAL NETWORK APPROACH TO PREDICT THE VALUE OF NEGOTIATION OF PROCUREMENT OF GOODS AND SERVICES - CASE STUDY PROCUREMENT OF GOODS AND SERVICES POWER PLANT AND ENERGY

• Main Author: NUGROHO NIM: 94516014, HERU
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/27851
• Institution: Institut Teknologi Bandung

Negotiation as one of the performance benchmarks of a procurement, which is consist of cost, quality and time. A good negosiator should be understand the knowledge of the negotiation itself, the negotiation variables, and the characteristics of the opposing negotiator. In WIKA, it is often difficult to get a reference to the magnitude of the value of negotiation due to variations and variable prices are very diverse. A formulation of the value of negotiation is necessary to facilitate the negotiator to evaluate the negotiations that will and have been done. By taking the negotiation variables from several sources, and then formulated with multi-variate regression and neural network, it is expected to obtain predictive prediction of negotiation value and prediction comparison when using multi-variate regression and neural network. <br /> <br /> This study takes the negotiation variable from Dzeng and Wang (2017), that from a survey ever conducted to 90 contractors in Taiwan, from 10 key variables of negotiation, there are 7 variables that are quite dominant. These variables include, quantity, price, mode of payment, down payment, delivery (schedule of implementation), duration of payment, and transportation. By taking the 7 variables in WIKA environment especially Power Plant and Energy Department, and using multi variate regression and neural network, it is expected before negotiation, a negotiator in WIKA environment has got a reference about the amount of negotiation value that must be done. <br /> <br /> From the analysis of predicted data, for multi variate regression obtained data TS = 89.0913, MAD 0.0722 and MSE 0.0129, while for neural network obtained data TS = 5,6905, MAD = 0,0748 and MSE = 0,0099. For tracking signal value, if the value is closer to zero or marked negative, this means the actual value is almost always smaller than the prediktive value. As for the value of MAD is relatively similar both for multi-variate regression and for neural network. And lastly the value of MSE generated on the neural network equation is smaller when compared to the MSE predicted results generated in the multi variate regression equation.