Visual Analytics (VA) supports iterative exploration and insight discovery, but current frameworks lack sufficient provenance management to capture the sequence of analytical actions and reasoning. This limitation hinders the reconstruction of analytical processes and the generation of coherent narratives. We propose the VASTorytelling Framework, which extends existing VA framework by incorporating mechanisms for recording user interactions, annotating visualizations, and utilizing extended data—including logs and external knowledge—for narrative construction. The framework supports metacognitive reflection and integrates large language models to assist narrative generation and reduce cognitive load. This approach enhances traceability, explainability, and communication in long-term VA activities.

This paper explores the challenges and solutions in graph visualization, a critical tool used across various fields for representing complex relationships between objects. As the complexity and volume of these relationships increase, the readability and effectiveness of graph visualization diminish. Individual techniques such as node layout and edge bundling have been employed to mitigate this issue, with methods like the Kamada-Kawai for node layout and Force Directed Edge Bundling for edge bundling showing success. However, these techniques are often conducted separately in network visualization. This paper discusses a method proposed by Meikari et al. that aims to improve visualization results by simultaneously conducting node layout and edge bundling. This method uses evolutionary computation to optimize both aspects at once, potentially preventing the decrease in effectiveness due to inappropriate node layouts or overemphasis on edge bundling. However, the execution time and validation of this method remain significant challenges. This paper aims to further explore and address these issues.

This paper proposes a method that includes the process of adjusting the position of nodes in edge bundling by integrating an edge bundling and a node layout into a single genetic algorithm. Edge bundling is one of the graph visualization methods that can improve complex graphs and produce visually superior graphs. However, most edge bundling don't include the process of adjusting the position of nodes. Our approaches includes the process of adjusting the position of nodes in edge bundling by integrating an edge bundling and a node layout into genetic algorithm. Then, we apply our method to graphs and evaluate the generated graphs to investigate the effectiveness of our method.

In recent years, various graph pooling methods have been proposed, and the existing edge pooling methods have some problems. Edge pooling aggregates nodes by removing edges while considering some node characteristics. However, edge pooling ignores the surrounding node features and graph topology. We propose a novel graph pooling method to address this problem. To address the problem, we build a reasonable pooling graph topology, consider the structure and feature information of the graph, improve the objectivity of node selection, and use the edge pooling method to select edges after considering the structure and feature information of the graph. Experimental results on the dataset show that our method is effective in graph classification and outperforms state-of-the-art graph pooling methods.

Hierarchical topic model is the method used in considering topics with hierarchical relationships. Neural embedding topic labelling (NETL) is a method utilized to label topics with neural embedding, even though it labels topics without topic relationships. The labels of hierarchical topics should have hierarchical relationship with other labels. This study proposes a method for labeling hierarchical topics with hierarchical relationships, and uses NETL to generate candidate labels for bottom topics. Moreover, our proposed method calculates how small the overlap of the candidate labels compared with other sibling topics. To label the upper topics, our method adds the label of the bottom topics and generate labels in the same way as the bottom topics recursively. Our method succeeded label hierarchical topics with labels which is more qualitative labels to consider hierarchical relationship of topics.

Hierarchical topic model is the method used in considering topics with hierarchical relationships. Neural embedding topic labelling (NETL) is a method utilized to label topics with neural embedding, even though it labels topics without topic relationships. The labels of hierarchical topics should have hierarchical relationship with other labels. This study proposes a method for labeling hierarchical topics with hierarchical relationships, and uses NETL to generate candidate labels for bottom topics. Moreover, our proposed method calculates how small the overlap of the candidate labels compared with other sibling topics. To label the upper topics, our method adds the label of the bottom topics and generate labels in the same way as the bottom topics recursively. Our method succeeded label hierarchical topics with labels which is more qualitative labels to consider hierarchical relationship of topics.

Recommender system is an information-filtering tool used in solving the problem that the user's preference in information overload. In recent years, some algorithms have been combined with some side information (i.e., item description documents, user reviews, and social networks), and rating prediction accuracy has been significantly improved. However, for fashionable goods, such as apparel and shoes that are important for designing, the contextual information of items is insufficient, and their image shape feature should be considered. Currently, no such recommender system is available to use this feature of image shape. This study proposes a novel probabilistic model using the image shape feature that integrates a convolutional neural network into the probabilistic matrix factorization. The experiment conducted on two real-world datasets corroborates that our model outperforms the other recommendation models.

In this paper, we propose a fuzzy markup language (FML)-based feature similarity assessment agent with machine-learning ability to evaluate easy-to-learn degree of the Japanese and Taiwanese words. The involved domain experts define knowledge base (KB) and rule base (RB) of the proposed agent. The KB and RB are stored in the constructed ontology, including features of pronunciation similarity, writing similarity, and culture similarity. Next, we calculate feature similarity in pronunciation, writing, and culture for each word pair between Japanese and Taiwanese. Finally, we infer the easy-to-learn degree for one Japanese word and its corresponding Taiwanese one. Additionally, a genetic learning is also adopted to tune the KB and RB of the intelligent agent. The experimental results show that after-learning results perform better than before-learning ones.

Numerous information visualization techniques are available for utilizing and analyzing large data. Among these techniques, network visualization, which employs node-link diagrams, can determine the relationship among multi-dimensional data. However, when data become extremely large, visualization becomes obscure because of visual clutter. To address this problem, many edge bundling techniques have been proposed. However, although graphs present several edge types, previous techniques do not reflect these edges. In this paper, we propose a new edge bundling method for multi-type co-occurring graphs. In this method, electro-static forces work between each pair of edges; however, if the edges are of different types, then repulsion works between pairs. By bundling edges of the same type, a graph can more clearly show relationships among data. Qualitative evaluation through questionnaires lead to useful knowledge, i.e., the proposed method improves bundling performance more extensively than other related work.

This paper describes a method for integration of multiple path models. Structural Equation Modeling (SEM) is useful for causality analysis. However, the result of SEM may lack reliability owing to the model being constructed upon subjective assumptions. To mitigate this condition, this paper proposes the integration of multiple path models by OR and probability concepts. Four integrations are performed in an experiment for each feature case using 13 path models and the usability of the methods is discussed.

This paper describes a case study of knowledge discovery from web access logs of an e-commerce site. It is important for e-commerce sites to analyze the behavior of visitors from web access log to increase sales and recurring users. In order to support the tasks, we use a visualization method both trend and relationships called FACT-Graphs with Sequential Probability Ratio Test for detecting trend change points. In an experiment for the data of a Web shop in Japan, we could extract 14 trend change points, visualize information both trend and relationships, and extract useful knowledge of the improvement of navigation for visitors.

This paper describes a causality analysis of best-selling software games. Software game makers typically distribute advance information on their products for advertising purposes, in the hope of earning favorable user reviews. Games highly rated by users typically become best sellers, but what information significantly affects their sales has not yet been analyzed quantitatively. We used data on software games to analyze sales factors with a multiple regression model and structured equation modeling (SEM). Experiment results using data obtained from webpages on 746 software games show that hardware types, specific makers, and game genres have a major motivational impact on customers' purchases.

A method for improving a visualized preference transition network by screening nodes in the network, where a node represents a product item, is described. The original preference transition network was developed not only for visualizing customer movements/trends in selecting items but also for finding the features of items. However, understanding such movements/trends and features is difficult when the network has many nodes and links. To solve this problem, the proposed method is a sensitivity analysis for identifying redundant nodes and links with adjustment of the threshold expressed by the Simpson coefficient. The effectiveness of this method was shown through a numerical experiment for 172 kinds of products, 2,227 customers, and their 90,000 sales records.

This paper proposes a hotel recommender system based on sales records. Basic premise under the research is that the sales records include the user's preference relations among hotels. The proposed system recommends hotels based on preference transition network when a user selects a hotel. This paper describes four steps procedure for building the preference transition network proposes in detail. The proposed recommender system is available for repeatable purchase without explicit product evaluation. The features of the prototype system are also illustrated.

This paper introduces a knowledge management system, called RESTER2 (Reusable ToDo Synthesizer 2). The goal of RESTER2 is to redesign existing processes, organize and share knowledge about organizational practices, and invent new processes. Further, RESTER2 manages ToDo which is instantiated from the stored process. This paper demonstrates the function of RESTER2 with process ontology and progress database, and it also presents the user interface. Comparing RESTER2 to the previous RESTER which has not ontology, this paper verifies the usefulness of the process ontology for synthesizing ToDo.

This paper proposes architecture for workflow cooperation among organizations. To design system architecture, business processes are classified and systematized. The proposed architecture consists of three components: request message converter, business process auto-generator and business process manager. Using WordNet for pair of words, the first component converts a message to a trigger for the second component. Then according to MIT Process Handbook for business process repository, the second component automatically generates a business process instance in an operating order. Finally, referring to case base, the third component decides and manages proper schedule and transactors for the instance from case-base. In an experiment, this paper also validates the proposed architecture and discusses its feasibility.

To recognize the trends in journal papers, this paper discusses a text mining method and its application. The method is based on combination of the conventional TF-IDF algorithm for document indexing and KIM analysis in marketing research. While TF (term frequency) can be clue for strength of topics and IDF (inverted document frequency) can be clue for bias of topics, recency in RFM analysis can be clue of vicissitude of topics. Applying the proposed method to trend analysis for the quality control journals in the Japanese society, this paper describes how the cross-tabulation of TF, DF and LA (last appearance) recognizes the research trends.

This paper describes the sales channels decision problem on balancing opportunity loss and supply surplus. Service industries which sell limited, perishable inventories such as flight tickets and hotel rooms through multi sales channels are constantly faced with the problem how to allocate their inventories for getting the optimum total revenues because a high profitable channel has weak sales power and a low profitable channel has strong one. In order to solve this problem timely and reasonably, this paper formulates an original problem as dynamic programming where the decision factors are induced from the past sales records. Achieving numerical simulation, this paper also shows the availability of the strategy with dynamic programming is shown.

In the real world, supply chains are too complex to analyze mathematically. Therefore, the concurrent simulator by object-oriented technology is a reasonable solution for the system analysis and the decision-making. This paper describes the SCM simulator which consists of participants named player and e-marketplace. In order to evaluate sensibility of a supply chain, this paper proposes to embed policy and capacity in players who are classified as follows: end customer, intermediate supplier, end supplier, and transportation server. For the first step, this paper introduces the prototype system for the case that the customer-supplier relation is established.

This paper describes e-marketplace that is a virtual trading area of SCM (supply chain management) simulator called LOSIMOPU. Being implemented by object oriented technology; LOSIMOPU allows players objects such as end customers and parts suppliers to trade in e-marketplace. The requirements for e-marketplace are as follows: (1) each player object can join in and exit from the area independently, (2) each player object trades concurrently, and (3) the status of each transaction can be monitored and recorded to databases for the analysis. To satisfy the requirements, this paper proposes (a) agent-type interface, (b) trading protocol, and (c) scalability. The prototype system based on these requirements is also introduced.