Kartick Chandra Mondal

A new approach for association rule mining and bi-clustering using formal concept analysis

Association rule mining and bi-clustering are data mining tasks that have become very popular in many application domains, particularly in bioinformatics. However, to our knowledge, no algorithm was introduced for performing these two tasks in one process. We propose a new approach called FIST for extracting bases of extended association rules and conceptual bi-clusters conjointly. This approach is based on the frequent closed itemsets framework and requires a unique scan of the database. It uses a new suffix tree based data structure to reduce memory usage and improve the extraction efficiency, allowing parallel processing of the tree branches. Experiments conducted to assess its applicability to very large datasets show that FIST memory requirements and execution times are in most cases equivalent to frequent closed itemsets based algorithms and lower than frequent itemsets based algorithms.

Prediction of protein interactions on HIV-1-human PPI data using a novel closure-based integrated approach

Discovering Protein-Protein Interactions (PPI) is a new interesting challenge in computational biology. Identifying interactions among proteins was shown to be useful for finding new drugs and preventing several kinds of diseases. The identification of interactions between HIV-1 proteins and Human proteins is a particular PPI problem whose study might lead to the discovery of drugs and important interactions responsible for AIDS. We present the FIST algorithm for extracting hierarchical bi-clusters and minimal covers of association rules in one process. This algorithm is based on the frequent closed itemsets framework to efficiently generate a hierarchy of conceptual clusters and non-redundant sets of association rules with supporting object lists. Experiments conducted on a HIV-1 and Human proteins interaction dataset show that the approach efficiently identifies interactions previously predicted in the literature and can be used to predict new interactions based on previous biological knowledge.

MOSCFRA: a multi-objective genetic approach for simultaneous clustering and gene ranking

Microarray experiments generate a large amount of data which is used to discover the genetic background of diseases and to know the characteristics of genes. Clustering the tissue samples according to their co-expressed behavior and characteristics is an important tool for partitioning the dataset. Finding the clusters of a given dataset is a difficult task. This task of clustering is even more difficult when we try to find the rank of each gene, which is known as Gene Ranking, according to their abilities to distinguish different classes of samples. In the literature, many algorithms are available for sample clustering and gene ranking or selection, separately. A few algorithms are also available for simultaneous clustering and feature selection. In this article, we have proposed a new approach for clustering the samples and ranking the genes, simultaneously. A novel encoding technique for the chromosomes is proposed for this purpose and the work is accompleshed using a multi-objective evolutionary technique. Results have been demonstrated for both artificial and real-life gene expression data sets.

SysML Based Conceptual ETL Process Modeling

Data generated from various sources can be erroneous or incomplete which can have direct impact over business analysis. ETL (Extraction-Transformation-Loading) is a well-known process which extract data from different sources, transform those data into required format and finally load it into target data warehouse (DW). ETL performs an important role in data warehouse environment. Configuring an ETL process is one of the key factor having direct impact over cost, time and effort for establishment of a successful data warehouse. Conceptual modeling of ETL can give a high-level view of the system activities. It provides the advantage of pre-identification of system error, cost minimization, scope, risk assessment etc. Some research development has been done for modeling ETL process by applying UML, BPMN and Semantic Web at conceptual level. In this paper, we propose a new approach for conceptual modeling of ETL process by using a new standard Systems Modeling Language (SysML). SysML extends UML features with much more clear semantics from System Engineering point of view. We have shown the usefulness of our approach by exemplifying using a use case scenario.

A complete review of computational methods for human and HIV-1 protein interaction prediction

Human Immunodeficiency Virus Type 1 HIV-1 has grabbed the attention of virologists in recent times owing to its life-threatening nature and epidemic spread throughout the globe. The virus exploits a complex interaction network of HIV-1 and human proteins for replication, and causes destruction to the human immunity power. Antiviral drugs are designed to utilise the information on viral-host Protein-Protein Interactions PPIs, so that the viral replication and infection can be prevented. Therefore, the prediction of novel interactions based on experimentally validated interactions, curated in the public PPI database, could help in discovering new therapeutic targets. This article gives an overview of HIV-1 proteins and their role in virus-replication followed by a discussion on different types of antiretroviral drugs and HIV-1-human PPI database. Thereafter, we have presented a brief explanation of different computational approaches adopted to predict new HIV-1-human PPIs along with a comparative study among them.

Comparative Analysis of Structured and Un-Structured Databases

The introduction of relational database systems helped in faster transactions compared to the existing system for handling structured data. However, in course of time the cost of storing huge volume of unstructured data became an issue in traditional relational database systems. This is where some unstructured database systems like NoSQL databases were introduced in the domain to store unstructured data. This paper focuses on four different structured(PostgreSQL) and un-structured database systems(MongoDB, OrientDB and Neo4j). In this paper, we will eventually see the different kind of data models they follow and analyze their comparative performances by experimental evidences.

Comparative Study of Parallelism on Data Mining

Today’s world has seen a massive explosion in various kinds of data having some unique characteristics such as high-dimensionality and heterogeneity. The need of automated data driven techniques has become a necessity to extract useful information from this huge and diverse data sets. Data mining is an important step in the process of knowledge discovery in databases (KDD) and focuses on discovering hidden information in data that go beyond simple analysis. Traditional data mining methods are often found inefficient and unsuitable in analyzing today’s data sets due to their heterogeneity, massive size and high-dimensionality. So, the need of parallelization of traditional data mining algorithms has almost become inevitable but challenging considering available hardware and software solutions. The main objective of this paper is to look at the need and limitations of parallelization of data mining algorithms and finding ways to achieve the best. In this comparative study, we took a look at different parallel computer architectures, well proven parallelization methods, and programming language of choice.

Brief Review on Optimal Suffix Data Structures

Suffix tree is a fundamental data structure in the area of combinatorial pattern matching. It has many elegant applications in almost all areas of data mining. This is an efficient data structure for finding solutions in these areas but occupying good amount of space is the major disadvantage of it. Optimizing this data structure has been an active area of research ever since this data structure has been introduced. Presenting major works on optimization of suffix tree is the matter of this article. Optimization in terms of space required to store the suffix tree or time complexity associated with the construction of the tree or performing operation like searching on the tree are major attraction for researcher over the years. In this article, we have presented different forms of this data structure and comparison between them have been studied. A comparative study on different algorithms of these data structures which turns out to be optimized versions of suffix tree in terms of space and time or both required to construct the tree or the time required to perform a search operation on the tree have been presented.

Closure Based Integrated Approach for Associative Classifier

Building a classifier using association rules for classification task is a supervised data mining technique called Associative Classification (AC). Experiments show that AC has higher degree of classification accuracy than traditional approaches. The learning methodology used in most of the AC algorithms is apriori based. Thus, these algorithms inherit some of the Apriori’s deficiencies like multiple scans of dataset and accumulative increase of number of rules. Closed itemset based approach is a solution to the above mentioned drawbacks. Here, we proposed a closed itemset based associative classifier (ACFIST) to generate the class association rules (CARs) along with biclusters. In this paper, we have also focused on generating lossless and condensed set of rules as it is based on closed concept. Experiments done on benchmark datasets to show the amount of result it is generating.

Knowledge discovery from HIV-1-human PPIs assimilating interaction keywords

Todays world is gradually getting agitated by Human Immunodeficiency Virus — Type 1 due to its pervasive and death-dealing nature. The virus replicates by exploiting a complex interaction network of HIV-1 and human proteins and destructs human immunity power, gradually leading to AIDS. Anti-HIV drugs are designed to utilize the information on viral-host protein-protein interactions (PPIs), so that the viral replication and infection can be prevented. In this article, we have presented an effective computational approach using pattern-mining based algorithm in order to predict novel interactions between HIV-1 and human proteins based on the experimentally validated interactions curated in public PPI database. Additionally, we have tried to provide the information on interaction types associated with each of the predicted interactions with an estimated confidence. Further, we have analyzed our predicted interactions by finding overlap with other studies. We believe that this article would enhance the discovery of HIV-1 medications in a way of designing the drug targets.