Designing dynamic forecasting system for truckload market pricing in logistics sector

The logistics sector is an ever-growing industry throughout the world. A robust logistics sector is one of the common features of industrialized countries. The main reason behind highway transportation being the major component of the logistics sector is that the sector is closely associated with other sectors. From this perspective price forecasting for the truckload spot market in the logistics sector has been an essential field of research recently. It has been an area of interest for decision makers and managers of companies to make predictions for the future in any sector in which uncertainty exists. Price forecasting has great importance, particularly in the case of decision making regarding a new area and an investment in economic terms. Price forecasting of the truckload spot market in the logistics sector is one of the critical areas of decision making for the determination of future value from the viewpoint of truckers. From this perspective price forecasting in the truckload spot market is especially important to predict price fluctuations in the short term and to protect third-party providers from potential risks and crises. Moreover, large companies can plan their budgets according to the price information. At this stage one of the essential points is to create a price forecasting model addressing all the variables affecting it, in which the customer and the trucker can find common ground. This study used two methodologies to determine truckers' spot price in the freight transport process, and it created a price forecasting framework. The first model created in the price forecasting system is a quantile regression, one of the econometric models. The second method used in the study is artificial neural network, an artificial intelligence method. Truckers' price varies depending on several factors, such as the type of transportation, departure and arrival cities, road conditions, type of route, amount of freight, and location difficulty level. The objective of this study is to develop a framework for forecasting the truckload spot market price in the logistics sector, to offer existing customers the most economical and high-quality services, to forecast the short- and long-term risks and price fluctuations that may be faced in the sector, and to design an effective model to determine the acceptable price levels of truckers. The designed price forecasting approach will ensure that the performance and forecast levels of both the artificial intelligence and the econometric model can be compared. In this scope two different price forecast models will be formed, the output performances will be assessed, the best methodology will be selected, and projections will be carried out. The main aim is to determine the price information dynamically, accurately, and in an expeditious and reliable manner, thus improving the quality and performance criteria in the truckload spot market, minimizing the uncertainties, risks, and errors that may be faced in the future, preventing price fluctuations, and ensuring the most economical and reliable decision support mechanism in the sphere of logistics and supply chains. From the managerial perspective, forecasting methods give projections that rely on the strength of past data, so business executives can obtain a mixture consisting of raw data and personal impressions to develop a useful forecast. This means that businesses can use the method to make decisions about their future. Managers can design possible future outcomes for their business. Different aspects of management can be considered, such as creating, developing, producing, and adding value for the organization. The paper has two objectives. Firstly, the price forecasting approach is developed with two methods. The quantile regression and the conventional back-propagation neural network (BP) models are used to forecast the price of highway transportation. The mean absolute percentage error (MAPE) is used to assess the forecast performance of the two methods. Secondly, the models are applied to two concepts: a route-based model and a general model in which all routes are assessed together. This thesis also investigates customer perspective of dynamic pricing for highway transportation. On the customer side, the problem that needs to be solved is to estimate the price that can be accepted and to calculate the customer bid response functions. Customer behavior can be modeled on this side and the probability of accepting the price to be given can be calculated. Customer is primarily segmented within this context. The logit model is used for the bid response functions. Through the model, detailed analysis and strategies have been defined for the customer. Price tendencies and probabilities have been obtained according to different customer segmentations. In the last part of the thesis, the price to be given to the customer is determined. For this purpose, two approaches have been determined. Determination of the price so as to maximize the profit to the company with the first nonlinear optimization. Models are analyzed both by uncontrained and by adding limit of market constraint of certain customer groups. The second approach is to estimate the price to be given to the customer. Linear regression ,artificial neural network and quantile regression method are used in this case. It is observed that the quantile regression gives better results when the selling price is determined. The results of optimization methods provide more fluctuating price information. However, firms can use the optimization model in their strategic decisions over a longer period of time. Forecast models have shown to give more similar results in the real situation when the instant sale price is determined. The aim of this thesis is to develop the decision support model for road transport in the logistics sector, to provide the most economical and high quality services to existing customers, to forecast short and long term risk and price fluctuations in the sector, model design. In this designed three-legged system, it is another purpose to provide profit maximization that the vehicle firm wants to achieve. In this context, a statistical and price forecasting model will be created and the selection process and prospective projections of the best methodologies and scenarios to evaluate output performances will be realized. By using the developed price forecasting approach system, uncertainty and risk factors can be eliminated and proper price information can be gained. Thus, an economical and sustainable supply chain network design can be achieved.