Treemap Layout Based on Monte Carlo Tree Search

Treemaps are a commonly used method for hierarchical data visualization.The previous research shows that three treemap layout algorithms(SQR,STR,SPI)that meet the principle of gradually adding points on four common optimization indicators(average aspect ratio,stability,consistent data,readability)have their own advantages and disadvantages.Both Strip and Spiral algorithms result in relatively large as-pect ratio of layouts,and the squarified treemap has some room for optimization in the aspect ratio.Based on Monte Carlo tree search,the corresponding search tree is constructed while laying out the treemap and the upper confidence bound formula is used as the selection criterion for the search path.For the three layout algorithms,the arithmetic average aspect ratio and the newly proposed weighted aspect ratio are improved within acceptable time for users.To show the effectiveness of the algorithm,the evaluation with the con-structed single-layer dataset which with 3 different distributions and 7 different scales and multi-layer real dataset is conducted and then the results are compared with the original algorithm.The guidelines for users to select a suitable layout algorithm and set proper parameters are also provided.