Takenao Ohkawa

Discovery of Quantitative Sequential Patterns from Event Sequences

In this paper, we consider the problem of frequent pattern mining in databases of temporal events with intervals. Since quantitative temporal information might play important roles in many application domains, it is critical to discover patterns to which numerical attributes are associated. To this end, we consider two kinds of temporal patterns with quantitative information on the durations and time differences of events, and propose corresponding algorithms by incorporating numerical clustering techniques into existing temporal pattern miners. The effectiveness of the proposed algorithms was assessed by using real world datasets.

Stochastic qualitative reasoning and its application to diagnosis of air conditioning system

The simulation technique on qualitative models is one of the powerful tools for the fault diagnosis. However, since a qualitative model is expressed with simple causal relations, enormous number of behaviors are possible even pruning off improbable behaviors by some real measured values patterns. Therefore, it is hard to apply it to practical scale system as it is. To cope with this problem, a stochastic qualitative reasoning is proposed in this paper. In this method, all of the behaviors have existence probabilities. The behaviors with relatively small probability are eliminated. We have applied this method to diagnosis of the air conditioning system. If a trouble is reported, the qualitative models of the target system in normal and failure are simulated and compared with the real measured values in the target plant. Agreement degrees with the model and the measured values are calculated with existence probabilities, and the cause of failure is found out. Through this application, we have confirmed the effectiveness of the method.<<ETX>>

A method of comparing protein molecular surface based on normal vectors with attributes and its application to function identification

Recent researches have clarified that the function of protein depends on its molecular surface. They suggest the possibility of the protein function identification based on the molecular surface comparison, in which a molecular surface of an unknown protein is compared with many surfaces of known active sites as reference templates. This paper presents an effective surface comparison method by using normal vectors with attributes of the curvature and the physical properties on projections and depressions. The vectors that should be matched are limited by extracting two vectors at similar relative positions and with the attributes of surface in order to reduce computational complexity. The proposed method was applied to 11 surface data. As a result, the mean calculation time was about 3 min, and it was possible to compare at approximately an optimal location. This method was applied to 103 surface data. The result of identification showed 95.2% identification accuracy.

Protein structure classification by structural transformation

Protein structure classification plays an important role in understanding the relationships among structure and sequence. Recently, as the number of known protein structure are increasing steeply, automatic classification is highly required. This paper presents a new method of classification of protein folds based on the structural similarities between a pair of proteins from the viewpoint of the secondary structural elements. The similarity can be evaluated through the structural transformation from one protein to the other, which is composed of a set of predefined primitive operations. Since the similarity is scored with the total cost of the transformation, the method requires no threshold value. The effectiveness of the method was empirically demonstrated through the experiments using the data in the Brookhaven Protein Data Bank.

Scenario generator for qualitative simulation system

A qualitative simulation system has been developed for understanding the behavior of a system modeled by a structural model. In decision making, it is necessary to understand the influence flow, namely how the model derives the behavior of the concerning node in the model rather than the behavior of each node. This paper presents a mechanism of scenario generating for explaining the influence flow to the concerning node. The mechanism consists of the following processes: 1) by tracing the structural model, all paths from the input node, which triggers the initial change of the system, to the concerning node are derived, and using similarity of behavior, the influence flow to the concerning node is selected out of all paths; 2) important nodes are selected according to the type of nodes; and 3) the scenario document is generated along the selected flow using the sentence template and the word dictionary.

Real-time stochastic qualitative simulation of large scale air conditioning systems

This paper gives a new algorithm for the efficient stochastic qualitative simulation, which can suppress the number of generated behaviors using the existing probabilities of states. In the new algorithm, the minimum existence probability, which is the lower bound of existence probability of states being preserved, is estimated with the Monte Carlo method. As a result, the states that are insignificant for envisioning useful behaviors are removed before generating them. In addition, the prefetch of measurement to check consistency of the states with the measurement in advance contributes to efficiency of the states generation. Through the simulation of the large scale building air conditioning systems, we clarify that the number of states can be reduced under computable order and the simulation terminates in real time.

Diagnosis of an air conditioning system by stochastic qualitative reasoning

A diagnosis using qualitative reasoning is applicable to the diagnosis of building air conditioning systems such that their configuration is various. In this technique, a target system is modeled qualitatively, and possible patterns of behaviour of the target model are generated by the qualitative reasoning. By comparing the results of reasoning with the real measured values, a faulty part is identified. We propose a stochastic qualitative reasoning method. In this method, the existence probability is introduced to prevent the increasing possible behaviours of the model. Through the qualitative reasoning process, the states having negligible probability are eliminated. We have applied this method to fault diagnosis of a building air conditioning system and confirmed its effectiveness.<<ETX>>

Frequent Pattern Discovery from a Single Graph with Quantitative Itemsets

In this paper, we focus on a single graph whose vertices contain a set of quantitative attributes. Several networks can be naturally represented in this complex graph. An example is a social network whose vertex corresponds to a person with some quantitative items such as age, salary and so on. Although it can be expected that this kind of data will increase rapidly, most of current graph mining algorithms do not handle these complex graphs directly. Motivated by the above background, by effectively combining techniques of graph mining and quantitative itemset mining, we developed an algorithm named FAG-gSpan for finding frequent patterns from a graph with quantitative itemsets.

Dispatching rule exchangeable optimization planning system

A new method for selection of dispatching rules, called a status selection method is discussed. In this method, the most promising status is selected from tentative statuses that are generated by applying all dispatching rules. Since the dispatching rules and the knowledge of status selection are independent, it is easy to exchange the dispatching rules. From the result of the application of the proposed method to a simple flow shop problem, it is confirmed that the status selection method is effective.<<ETX>>

Real-time simulation for fault detection and diagnosis using stochastic qualitative reasoning

Qualitative reasoning is one of the fault detection and diagnosis methods and is an effective way for complex systems. Especially, in the stochastic qualitative reasoning, the probabilistic process is used for state transition of the qualitative models with probabilities. In the paper, the stochastic qualitative reasoning is applied to a modelling method of the human qualitative thinking process based on the instrumentation diagram and the control information in building air conditioning systems. However, several subjects remain to be solved for practical uses. The paper proposes real-time simulation using stochastic qualitative reasoning as a way of practical uses for fault detection and diagnosis of a VAV (variable air volume) system of air conditioning systems.