Vietnam Journal of Computer Science

Efficient operation of the integrated management information systems (IMISs), especially multi-agent systems, is related to their ability to automatically process collective knowledge. On the basis of this knowledge the decision-making process is realized in the business organizations. This paper presents issues related to framework for acquiring and acquisition subprocesses in a collective knowledge of business organization processing in IMIS. The main novelty of the developed framework is the coverage of all the areas of operation of an organization. Additionally, the inter-area knowledge for automatic strategic-level decision-making has been taken into consideration. The main improvements of this framework are that it allows for processing of the whole collective knowledge of business organization and it can be directly implemented in IMIS.

This paper concentrates on the newly improved algorithm for solving the single-machine total weighted tardiness (SMTWT) problem, called Algorithm Inspired by Tree and Sorting (AIT-S). The algorithm is based on searching the solution space along with the tree and sorting rules. The properties of the algorithm are studied taking into account the results of experiments made using the designed and implemented experimentation system. This system allows in tuning the parameters of the algorithm as well as to compare the effects obtained by AIT-S algorithm with effects of AIT algorithm, and known meta-heuristic algorithms. The paper shows that the proposed algorithm still requires some improvements, however, it seems to be promising.

In this research, we exploit a novel approach for propagation processes on a network related to textual information by using topic modeling and pretopology theory. We first introduce the textual agent’s network in which each agent represents a node which contains specific properties, particularly the agent’s interest. Agent’s interest is illustrated through the topic’s probability distribution which is estimated based on textual information using topic modeling. Based on textual agent’s network, we proposed two information diffusion models. The first model, namely Textual-Homo-IC, is an expanded model of independent cascade model in which the probability of infection is formed on homophily that is measured based on agent’s interest similarity. In addition to expressing the Textual-Homo-IC model on the static network, we also reveal it on dynamic agent’s network where there is transformation of not only the structure but also the node’s properties during the spreading process. We conducted experiments on two collected datasets from NIPS and a social network platform, Twitter, and have attained satisfactory results. On the other hand, we continue to exploit the dissemination process on a multi-relational agent’s network by integrating the pseudo-closure function from pretopology theory to the cascade model. By using pseudo-closure or stochastic pseudo-closure functions to define the set of neighbors, we can capture more complex kind of neighbors of a set. In this study, we propose the second model, namely Textual-Homo-PCM, an expanded model of pretopological cascade model, a general model for information diffusion process that can take place in more complex networks such as multi-relational networks or stochastic graphs. In Textual-Homo-PCM, pretopology theory will be applied to determine the neighborhood set on multi-relational agent’s network through pseudo-closure functions. Besides, threshold rule based on homophily will be used for activation. Experiments are implemented for simulating Textual-Homo-PCM and we obtained expected results. The work in this paper is an extended version of our paper [T. K. T. Ho, Q. V. Bui and M. Bui, Homophily independent cascade diffusion model based on textual information, in Computational Collective Intelligence, eds. N. T. Nguyen, E. Pimenidis, Z. Khan and B. Trawiski, Lecture Notes in Computer Science, Vol. 11055 (Springer International Publishing, 2018), pp. 134–145] presented in ICCCI 2018 conference.

Choosing the best schools from a group of schools is a multi-criteria decision-making (MCDM) problem. In this paper, we have represented a method that uses the fusion of two multi-criteria decision-making methods, Best–Worst Method (BWM) and Analytic Hierarchy Process (AHP), to rank some of the user preferred alternatives. The system considers the choice of the user and the quality of the alternatives to rank them. User preferences on the criteria are taken as inputs in the form of best–worst comparison vectors to measure the choice of the user. These values are applied to calculate the numeric weights of each of the criteria. These weights reflect the preference of the user. A dataset of secondary schools in Bangladesh has been compiled and used for automatic quantitative pairwise comparison on the alternatives to calculate the score of each alternative in every criterion, which reflects its quality in that criterion. These scores are calculated using AHP. The weights of the criteria as well as the scores of these alternatives in those criteria are then used to calculate the final score of the alternatives and to rank them accordingly. An extensive experimental analysis and comparative study is reported at the end of this paper.

There are several issues compromising the educational role of social networks, particularly in the case of video-based online content. Among them, individual (cognitive and emotional), social (privacy and ethics) and structural (algorithmic bias) challenges can be found. To cope with such issues, we propose a recommendation system for online video content, applying the principles of sustainable design. Precision and recall in English were slightly lower for the system in comparison to YouTube, but the variety of recommended items increased; while in Spanish, precision and recall were higher. Expected results include fostering the adoption of complex thinking by taking on account a user’s objective and subjective contexts.

In this paper, we propose a novel approach for transforming financial time-series values into symbolic representation based on value changes. Such approach seems to have few advantages over the existing approaches; one of the most obvious is noise reduction in the data and another one is possibility to find patterns which are universal for investigating different currency pairs. To achieve the goal we introduce a preprocessing method that allows initial data transformation. We also define a text-based similarity measure which can be used as an alternative for methods allowing to find the exact patterns in the historical data. To effectively evaluate our method, we present a concept that allows to predict the potential price movement direction and compare it with the actual price direction observed in the historical data. Such a method gives an opportunity not only to indicate the different price patterns based on the symbolic representation but also at the same time evaluate the predictive power of such patterns. The proposed approach is experimentally verified on 10 different currency pairs, each covering approximately a period of 10 years.

Recommender systems help users find relevant items efficiently based on their interests and historical interactions with other users. They are beneficial to businesses by promoting the sale of products and to user by reducing the search burden. Recommender systems can be developed by employing different approaches, including collaborative filtering (CF), demographic filtering (DF), content-based filtering (CBF) and knowledge-based filtering (KBF). However, large amounts of data can produce recommendations that are limited in accuracy because of diversity and sparsity issues. In this paper, we propose a novel hybrid method that combines user–user CF with the attributes of DF to indicate the nearest users, and compare four classifiers against each other. This method has been developed through an investigation of ways to reduce the errors in rating predictions based on users’ past interactions, which leads to improved prediction accuracy in all four classification algorithms. We applied a feature combination method that improves the prediction accuracy and to test our approach, we ran an offline evaluation using the 1M MovieLens dataset, well-known evaluation metrics and comparisons between methods with the results validating our proposed method.