Anjali Mahajan

Serial multimethod Combined Mining.

Combined Mining is an approach to combine various mining techniques to get more understandable and useful patterns from complex data. Different classical data mining techniques have their advantages and disadvantages so a single technique cannot be applied on different business data. Serial Multi-method Combined Mining (SMCM) is an approach where different mining techniques are used to get the patterns and finally the resultant patterns or rules are actionable. The actionable patterns are descriptive and assist us to finalize business decision. In SMCM, different mining methods are applied in predetermined sequential pattern. The resultant patterns of the previous method are also considered as a part of the input for the next method to be executed. SMCM helps to get advantages of different classical mining techniques to generate combined patterns but it needs the domain knowledge of business data for selection of different methods. The SMCM for credit card data by combining clustering and association techniques is demonstrated and experimental results are taken.

Domain driven approach for coherent rule mining

In Association Rule Mining, minimum support threshold is used to get the association rules. Deciding this threshold is quite a difficult task and has a great influence on the number and the quality of association rules. There is no chance of neglecting minimum support threshold as the large number of association rules generated missed some interesting rules discovered. The process of decision making with these rules may lead to undesirable and unexpected results. Minimum support threshold thus played a vital role in the entire process. To remove this dependency on minimum support threshold, we have proposed a framework which contains domain knowledge method, feature selection method and pruning technique to reduce the complexity of coherent algorithm to discover interesting positive and negative rules for business which are discovered based on the properties of propositional logic and thus do not require the minimum support threshold. In the initial part of the paper, we have explained the formation of coherent rules. Later, to reduce the complexity and make it more efficient we have added the feature of domain-driven to the framework of coherent rules and this feature is demonstrated with the help of implemented example. Further we have also introduced the concept of Combined Rule Mining which further enhances the results generated.

Genetic Algorithm-Based Approach for RNA Secondary Structure Prediction

In recent years, bioinformatics has become an essential subject for molecular biological study. The various available algorithms are used for analyzing and integrating biological data. Among many biological statistics RNA (ribonucleic acid) is one of the most important as it is used in protein synthesis. In computational molecular biology, the optimal secondary structure prediction of large RNA is a problem being faced today. RNA sequences of some virus are very large in number which requires a large amount of time for secondary structure prediction. Consequently, parallelization of algorithm is one of the solutions to diminish time consumption. This paper proposes the algorithm GAfold for predicting secondary structure of RNA on shared memory multicore architecture. The various RNA sequences as an input have been taken from Gutell database. For calculating minimum free energy, thermodynamic model is used and the outcomes are compared with existing algorithms.

Schedulability Analysis for Multiresource Scheduling in Middleware for Distributed Real Time Systems

Distributed Real Time Systems operate in resourceconstrained environments and are composed of tasks that must process events and provide soft real-time performance. The main function of a computing system is to provide services to its users. In order to perform its function, a computing system uses various resources such as processors, memory, communication channels, etc. Managing and scheduling these resources is an important function.

A semantic network approach to affect analysis: A case study on depression

This paper discusses the semantic network approach to identify affects in natural language input and focusses on representing spatio-temporal affect information. It has been observed that this approach performs better in analysis of affect information that can be effectively utilized for the prognosis of human cognitive behavior. The research work in this paper describes a new approach towards simple representation of multi-dimensional affect data that facilitates the provision of emotions or affects with varying granularity. The analysis algorithm thus executed on the semantic representation generates temporally significant behavior patterns. The analysis of the semantic network generates temporally significant behavior patterns. The framework has been designed to be extensible over a wide variety of applications in cognitive computing.