In e-commerce environments, the reputation of sellers is a crucial issue to potential buyers in making decisions. Most existing trust evaluation models compute a single value to reflect the general trust level of a seller without taking any transaction context into account. With such a simple trust evaluation model, a buyer may be easily deceived by a malicious seller in a transaction where the notorious value imbalance problem is involved, namely, the malicious seller can accumulate a high level of trust by selling cheap products and then start to deceive buyers by inducing them to purchase expensive products. In this article, we first present a trust vector consisting of four trust values (termed as CTT values), the computation of each of which is based on both past transactions and the forthcoming transaction. In addition, in the computation of CTT values, the parameters, such as product category, price range, and time range, can be specified and adjusted by the buyer, yielding different sets of trust values that can outline the seller's reputation profile. As a result, the value imbalance problem potentially existing in a forthcoming transaction can be identified. The computation of CTT values requires the pre-computation of aggregates over large-scale ratings and transaction data with necessary combinations of three dimensions (i.e., product category, price, and time), so as to deliver prompt responses to a buyer's query on different CTT values. In solving this challenging problem, we propose three new data structures and algorithms that mainly focus on efficient computation of CTT values. Our experiments conducted on both eBay data sets and large-scale synthetic data sets illustrate the advantages and disadvantages of our proposed approaches when responding to a buyer's CTT queries.
- Contextual transaction trust
- Data warehouse