Data Mining Driven Agents For Predicting Online

Auction markets provide centralized procedures for the exposure of purchase and sale orders to all market participants simultaneously. Online auctions have effectively created a large marketplace for participants to bid and sell products and services over the Internet. eBay pioneered the online auction in 1995. As the number of demand for online auction increases, the process of monitoring multiple auction houses, picking which auction to participate in, and making the right bid become a challenging task for the consumers. This project studies clustering based method is used to forecast the end-price of an online auction for autonomous agent based system. In the proposed model, the input auction space is partitioned into groups of similar auctions by k-means clustering algorithm. The recurrent problem of finding the value of k in k-means algorithm is solved by employing elbow method using one way analysis of variance (ANOVA). Then k numbers of regression models are employed to estimate the forecasted price of an online auction. Based on the transformed data after clustering and the characteristics of the current auction, bid selector nominates the regression model for the current auction whose price is to be forecasted.

Introduction

Predicting the end price depends on many factors, such as item type, type of auction, quantity available, opening price, number of bidders, average bid amount and many more. Price dynamics of the online auctions can be different even when dealing with auctions for similar items.

Functional data analytical tools have been used to characterize different type of auctions that exhibit different price dynamics. This project proposes a clustering based approach is used to characterize different type of auctions. The proposed model, the input auctions are clustered into groups of similar auctions based on their characteristics using k-means algorithm. To decide the value of k in k-means algorithm is a recurrent problem in clustering and is a distinct issue from the process of actually solving the clustering problem. The optimal choice of k is often ambiguous, increasing the value of k always reduce the error and increases the computation speed. In this paper, we are exploring Elbow approach using one way analysis of variance (ANOVA) to estimate the value of k. The end price for each cluster is estimated by employing distinct regression model for each cluster.

Bid selector nominates one of the multiple regression models based on the transformed data after clustering and the characteristics of the current auction whose end price is to be predicted. Then the nominated model predicts the end price of the auction.

Proposes System

Problem Study

A clustering based method is used to forecast the end price of an online auction for autonomous agent based system. In the proposed methodology the input auctions are partitioned into groups of similar auctions depending on their different characteristics. This partitioning has been done by using k-means clustering algorithm. The value of k in k-means algorithm is determined by employing elbow method using one way analysis of variance (ANOVA). Based on the transformed data after clustering and the characteristics of the current auction, bid selector calculates the end-price of the auction by employing regression model.

Literature Survey

There exist a large number of eBay automation tools, as well as a substantial body of research in price estimation, but the two fields appear to be largely separate. It seems that the eBay trading community does not realize the benefits that might be available to it, while computer scientists and economists are more interested in theoretical aspects of price prediction. We would like to think that our system may be the first that bridges the gap between the two.

Common eBay trading tools include snipers and simple monitors which are capable of alerting the buyer of all items that match some static search criteria; these tools do not attempt to decide whether an item is a good buy or not. There are a number of studies analyzing factors that affect t the final price of an online auction, including the effect of particular attributes on final price, the artificial domain Trading Agent Competition (Wellman et al. 2004), explicit price prediction for online auctions using machine learning on data extracted from item titles for very narrow domains (Ghani and Simmons 2004; Ghani 2005), and agent-based median price prediction (Gregg and Walczak 2004).

The work most related to ours improves upon these earlier approaches by mining product descriptions and using boosted machine learning to predict the final price (Heijst, Potharst, and Wezel 2008). It achieves much better generality than earlier approaches and is empirically applied to entire product categories (Canon digital cameras and Nike men shoes). While their approach is more general than ours (they do not have to explicitly specify category-specific features to be extracted), our application performs better on the category of our choice. Their Mean Relative Error (MRE) is 34% on Nike shoes and 58% on Canon cameras. For comparison, we use the same metric to measure the accuracy of our results and achieve an MRE of 16% for laptops.

Efficient E-Commerce Agent Data

It proposes Data mining based multi-agent system has been designed in favor of a multiple on-line auctions environment for selecting the auction, in which the traded item will be sold at the lowest price. The K-means clustering technique has been used to classify auctions into discrete clusters. Clustering operates dynamically on multiple auctions as bid price changes in running auctions. The results of the dynamic clustering are fed into the agents, and by employing probability-based decision making processes, agents deduce the auction that is most likely to close at the lowest price. Experimental results have demonstrated the robustness of the designed system for multiple on-line auctions with little or no available information

Prediction end-price of online auctions

Predicting end price of an online auction has been stated as a machine learning problem and has been solved using regression trees, multi-class classification and multiple binary classification. Among these machine learning techniques, posing the price prediction as a series of binary classification has been proved to be the best suited method for this task.

Assessing the Accuracy of Grey System Theory against Artificial Neural Network in Predicting Online Auction Final Price

The development of a predictor agent that utilizes Grey System Theory to predict the on-line auction final price in order to maximize the bidder’s profit is presented. The performance of this agent is compared with an Artificial Neural Network Predictor Agent (using Feed-forward Back-propagation Prediction Model). The work claims to be better than fuzzy or artificial neural network model. It considers few final prices in the past but does not consider other factors like bid rate, number of bidders or concurrent auctions etc., that influence the final price. All auctions are assumed to have begun at the same time and all the bidders participate from the beginning. Hence it does not consider the real world situations of the on-line auctions. This work is considered for comparison with our proposed model for predicting final prices.

Optimal Design of English Auctions with Discrete bid levels

The work presented considers only the log of the final price and does not consider parameters like the number of bidders, bid rate etc., which may not lead to accurate prediction. Large amounts of historical exchange of data from possible auction sites are collected, and machine learning algorithms are used in combination with the traditional statistical methods to forecast the final prices of products. An attribute construction method to overcome the problem of dynamically changing bid list is used.

A Simulation-Based Model for Final Price Prediction in Online Auctions

The authors have presented a simulated environment to test the proposed algorithms. The simulated results are further used to decide the final prices. By mimicking the bidder’s behavior, the prices are predicted by running the auctions in much iteration. The environment considered does not exactly mimic the real world situation as the bidder behavior and other changing factors are not considered.