Sushama Nagpal

Quality metrics for conceptual models for data warehouse focusing on dimension hierarchies

Multidimensional conceptual models have been accepted as the foundation for data warehouse designs. The quality of these models have significant effect on the quality of data warehouse and hence, in turn on the information quality. Few researchers have defined quality attributes for the conceptual models for data warehouse and have also proposed metrics to assess the quality attributes of these models objectively. The objective of this work is to propose candidate metrics to compute the structural complexity of multidimensional model. The main emphasis of this paper will be on the dimension hierarchies in multidimensional model. Though, these hierarchies play very significant role in analysing data at various granularity levels, their use enhances structural complexities of multidimensional model which can affect their understandability and modifiability and in turn maintainability.

Assessment of quality of data warehouse multidimensional model

Data warehouses are large repositories designed to enable the knowledge workers to take better and faster decisions. Due to its significance in strategic decisions, there is a need to assure data warehouse quality. One of the factors affecting the data warehouse quality is multidimensional model quality. Although there are some useful guidelines for designing good multidimensional data models, but objective indicators, i.e., metrics are needed to help designers to develop quality multidimensional models. Few researchers have proposed quality metrics for multidimensional models for data warehouse. These metrics need to be theoretically as well as empirically validated in order to prove their practical utility. In this paper, empirical validation using controlled experiment is carried out. We not only evaluate the effect of individual metric but also evaluate the effect of various combinations of metrics on data warehouse model quality specifically understandability, in order to best explain the variance of dependent variable due to independent variables. The results show that these metrics may be used as objective indicators for understandability. Finally, accuracy of our model in predicting the multidimensional model quality is also evaluated.

Complexity metric for multidimensional models for data warehouse

Quality of data models for data warehouse has significant effect on the quality of data warehouse. Complexity metrics play significant role in predicting quality attributes of a software artifact. Few researchers have proposed structural complexity metrics for the multidimensional data models for data warehouse which may act as objective indicators of the quality of these models. However, the metrics proposed earlier have not considered the structural complexity due to relationships among various elements present in these models. This paper proposes a complexity metric which considers structural complexity due to relationships among elements present in multidimensional models for data warehouse. The metric is proposed on the basis of Goal Question Metric approach. The practical usefulness of the proposed metric is proved by validating the metric using a practical framework proposed by Kaner. This preliminary validation suggests that the metric may be linked to the quality of the multidimensional models. The advantage of the metric is that it is available during early phase of software development life cycle. The metric will also help the developers to select quality data model among various semantically equivalent models.

Predicting quality of data warehouse using fuzzy logic

Due to strategic importance of data warehouse (DW) as decision support systems, it has become crucial to guarantee that these repositories should provide quality information to the decision makers. Quality of data warehouse multidimensional model has significant effect on data warehouse quality and in turn on the information quality. Few authors have suggested metrics to assess the quality of data warehouse multidimensional models. Empirical validation using statistical techniques like correlation analysis, univariate and multivariate regression techniques, etc., indicated that these metrics are significantly related to the quality of multidimensional models for data warehouse. But these techniques are not able to model non-linear relationship between the metrics and quality of multidimensional model. In this paper, model based on fuzzy logic approach is proposed to approximate non-linear relationship between the metrics and the quality of multidimensional models. In order to empirically evaluate the effectiveness of the proposed approach, validation is done on the published data and results indicate that the proposed model is able to predict the output with significant accuracy.

Real-time geo influence in social networks

Exponential burst of real-time information generated through Social Media Networks in recent years, creates a hot ubiquitous platform for research among data scientists. Main area under spotlight in Social Networks Analysis (SNA) is friendship networks, user influence and computation of how deep and how fast the information diffuses. Amidst recent upsurge in Smartphone usage statistics, there is spectacular rise in tagging of feeds and post them with locations due to GPS localizations (Geo-tags and location coordinates of images and news feeds). For the local advertisers and social campaigners, it is vital to carve out phase based user reach models in a particular city/location. In this paper, an algorithm to calculate real-time geo influence of twitter users pertaining to location and time constraints is proposed and necessitate its requirement to meet the vital requisite to find out how fast information is spread by a particular user in a certain city/area and hence providing a platform for ranking them based on their influence in a particular place, and measuring speed of information travel. Subject Descriptors-Data mining and social computing, Social and Behavioural Sciences.

Gravitational Search Algorithm in Recommendation Systems

Recommendation Systems have found extensive use in today’s web environment as they improve the overall user experience by providing users with personalized suggestions. Along with the traditional techniques like Collaborative and Content-based filtering, researchers have explored computational intelligence techniques to improve the performance of recommendation systems. In this paper, a similar approach has been taken in the form of applying a heuristic based technique on recommendation systems. The paper proposes a recommendation system based on a less explored nature-inspired technique called Gravitational Search Algorithm. The performance of this system is compared with that of a system using Particle Swarm Optimisation, which is a similar optimisation technique. The results show that Gravitational Search Algorithm excels in improving the accuracy of the recommendation model and also surpasses the model using Particle Swarm Optimization.

An empirical analysis of implicit trust metrics in recommender systems

Recommender system is an intelligent solution to information overload problem. Classical collaborative filtering based recommender system suffers from cold start and data sparsity problems. Incorporation of trust in classical recommender systems has potential to improve the overall performance of recommender system. Trust has been enormously researched and its influence is manifested in recommender systems. Because of unavailability of explicit trust information, various implicit trust metrics are developed to deduce trust from users online behavior. In this paper, we have conducted an empirical study of six implicit trust metrics on two different real world datasets. A comparative analysis of these metrics with classical user based collaborative filtering is performed.

Using Strong, Acquaintance and Weak Tie Strengths for Modeling Relationships in Facebook Network

Predicting strength of a relationship (also known as Tie Strength Problem) has been a trivial research area amongst sociologists for decades. However, considering the recent trends in internet behavior of people along with the development of so called social web, makes it popular amongst web scientists to work on this as a potential research topic with new perspectives. Real life is a complex social dynamic system comprising individuals starting of either as strong acquaintances or weak acquaintances and move towards strong or weak ties with passage of time. In this paper we validate the existence of varying degree of relationship individuals have on Facebook using unsupervised machine learning techniques like divisive hierarchical clustering and statistical techniques like SSE ; analyzing strength of the boundaries that distinguish them. We have realized this on a feature rich dataset of more than 100 nodes collected during 10th of July, 2011 to the 9th of September 2011 using a Facebook application. We provide descriptive error analysis interviews focussing on the clustered structure, obtaining it with an accuracy of 90%. The paper concludes by illustrating how modeling tie strength can improve social media design elements, including privacy controls, message routing and information prioritization in databases. Potential usage of this work can be in making complex recommender systems, lead generation marketing and in organizational or telecom network.

Empirical Validation of Object Oriented Data Warehouse Design Quality Metrics

Data warehouses have been developed that stores information enabling the knowledge worker to make better and faster decisions. As a decision support information system, a data warehouse must provide high level quality of data and quality of service. Various metrics have been defined and theoretical validated to measure the quality of the data warehouse in a consistent and objective manner and if quality measured, it can be managed and improved. Now, in this paper we will use these design quality metrics and empirically validated these metrics by conducting an experiment using regression analysis and deriving the conclusions according to the analysis so that they can be used by researchers and users.

Empirical investigation of metrics for multidimensional model of Data Warehouse using Support Vector Machine

Data Warehouse is the backbone of all analytics oriented organizations where business decisions need to be taken. Due to its role as a decision support system, its quality becomes crucial. Data warehouse conceptual models can be used to determine its quality during the early stages of design. Several metrics have been proposed to estimate the quality of these models. In order to corroborate the practical applicability of these metrics, it is important to validate them empirically. A number of propositions have been made in the past for the empirical validation of these metrics largely using statistical techniques of correlation and regression. However, statistical techniques are unable to model complex and non-linear relationships between the metrics and quality of the data warehouse models. In this paper, we have made an attempt to assess the non-linear relationship between the data warehouse structural metrics and understandability of its models by using Support Vector Machine (SVM). The results indicate that the proposed SVM model may aid in determining the understandability and inturn quality of the data warehouse conceptual models with high accuracy.