Recommendation system for popular tourist attractions in Taiwan using Delphi panel and repertory grid techniques

Problems exist in current on-line information systems available to tourists in that they are often of a general nature not offering a personalized service. To overcome this, in the context of the leading 100 Taiwanese attractions, a system is proposed comprising: (a) a construct elicitation mechanism using the Delphi panel method; (b) a matrix construction mechanism using the repertory grids, expert knowledge, and user preference; and (c) a recommendation mechanism using two similarity methods to provide recommendations. A study was consequently undertaken, and the main findings included: (1) the mean levels of precision in matching attractions with preferences were 60.4–79.3%, exceeding levels previously reported and so suggesting that the expected outcome of combining the Delphi panel method and repertory grid technique was achieved and (2) after decreasing the number of constructs, reductions in data-sparsity effect and dimensionality were also observed.