An Empirical Investigation of Properties of Different Random Forests in Dealing with Different Types of Data

Qiang Wan1*, Baoxun Xu1, Joshua Zhexue Huang2, Di Ma2, and Yunming Ye1
1Department of Computer Science, Harbin Institute of Technology Shenzhen
Graduate School, Shenzhen 518055, China
2 Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences,
Shenzhen 518055, China


Different random forests can be implemented by the integration of different training data sampling methods, subspace sampling methods, and tree building algorithms. This paper presents an empirical study on the performance of four random forests to reveal their distinct properties in handling real-world data of different characteristics. We first conducted a series of experiments to show how different random forests performed in terms of five statistical tests (strength, correlation, error bound, out-of-bag accuracy, and test accuracy) with changing feature subspaces. Then, we adopted some advantageous random forest models to solve three challenging classification problems: classifying very high dimensional data, classifying mixed type data, and classifying unbalanced data. We have observed that, for subspace sampling methods, random forests generated by Breiman's method outperformed those by Ho's method, whereas, for tree building algorithms, random forests generated by C4.5 outperformed those by CART For high dimensional data classification, a large subspace rather than Uoglm)+1J proposed by Breiman should be adopted to build a random forest model with high accuracy. Finally, random forests generated by the combination of both Breiman's sampling method and C4.5 algorithm outperformed Neural Network and Linear Discriminative Analysis methods in classifying unbalanced data.

Key words: Data mining, Machine learning, Ensemble classification, Random forests, Decision tree.

CREDIT-A Real Time Commit Protocol

Vishnu Swaroop, A. N. Singh, A. Anand, Udai Shanker* and S. Agrawal
Madan Mohan Malviya engineering College, gorakhpur-273010, India


This paper proposes a real-time commit protocol for distributed real time database systems (DRTDBS), Conflict Resolution via Elemental Dependency Inversion in Time (CREDIT) commit protocol. Here, two new dependencies are defined; one is Commit-on- Termination external dependency between final commit operations of lender and shadow of its borrower and other one is Begin-on-Abort internal dependency between shadow of borrower and borrower itself. If there is a serious problem in commitment of lender, execution of borrower is started with its shadow after aborting it and abort dependency created between lender and borrower due to update-read conflict is reversed to commit dependency between shadow and lender with read-update conflict. The final commit operations of lender and shadow of its borrower is governed by Commit• on- Termination dependency. Also, the borrower's shadow sends YES-VOTE message piggy bagged with the new resultto its coordinator in case of abort of lender & its borrower and activation of transaction's execution with help of shadow. /I minimizes the number of messages needed for execution and commit of cohort, and is also free from long duration locking of data items as in case of single phase commit protocol. The performance of CREDIT is compared with shadow PROMPT, SWIFT and DSS-SWIFT commit protocols for both main memory resident and disk resident databases with and withoul communication delay. Simulation results show that the proposed protocol improves the system performance up to 5% as transaction miss percentage. Also, it resolves the priority inversion problem of DDCR.

Keywords: Distributed Real Time Database System, Commit Protocol, Conflict Resolution, Dependency Inversion, Lender, Borrower.

Relevance of Business Intelligence Software-as-a-Service to Small Scale Industries

Preeti Singh*, Deepa Gaikwad, Archana Chimankar, Harshil Shah and Pranjal Srivastava
Symbiosis Center for Information Technology, Pune, India,

harshilshah in, pranjal.iet@gmail. com


Business Intelligence (BI) has always been around, helping organizations in managerial decision making, from Microsoft Excel being the simplest of all tools to open source and licensed vendors, offering a plethora of business centric features. In spite of the numerous varieties of tools available, BI has eluded the small scale industries (SS/s) in India. These organizations use traditional ways of decision making and analyze large amount of dormant information by transactional software, as BI remains unaffordable to them. By delving into the need of BI SaaS in the present scenario, its implementation, a comparative analysis of various vendors, coupled with inputs from industry experts enables one to form a better understanding of why SS/s are reluctant and how can they be helped by using SaaS as a platform to make BI a cost effective option.

Keywords: On Demand, Software as a Service (SaaS), Business Intelligence (BI), Small Scale Industries (SS/s), Vendor Analysis, Research/ Technology.

Mobile Money Order: The Next Step in Improving the Efficiency of Indian Postal Money Order

Ashok Bahadur Singh1* and Madhvi Sethi2
1Symbiosis International University, Pune, India.
2Symbiosis Institute of International Business, Pune, India



The growth of mobile banking in developing countries started capturing the attention of the researcher's business entrepreneurs, planners and governments for better utilization and extension of services to a large section of society. The research in mobile banking applications and its usability is growing very fast. The rapid transition and rollover of banks' branches towards core banking solution is yielding the transformation, evolution and integration of different entities to collaborate and integrate the business strategy on the various fronts such as micro finance institutions, postal services, insurance, telecom service providers etc. This paper is an attempt to study the mobile money order evolution, present status of Indian Postal Money Order, analyse the trend and suggest the approaches for increasing the efficiency of Indian Postal Money Order by using mobile banking as medium to integrate banking sectors with India Post. The proposed mobile money order for India Post would enhance the efficiency of present money order in terms of number of factors such as business growth, convenience, reduced transaction costs, social and economic developments and many more.

Keywords: Migrant remittances, Mobile Banking, Payment Systems, Money Orders, E-banking.

Understanding the MAC Layer Capacity of IEEE 802.11n

Shashank Tadakamadla*, Kaustubh Phanse and Md Sohail Ahmad
AirTight Network, Inc., Pune, India,
Email: {shashank. t, Kaustubh.phanse, md.ahmad}


IEEE 802. 11 n is a ratified standard for wireless local area networks that promises higher data rates and longer range. This article studies the throughput capacity of 802.11n medium access control layer. The emphasis is on the following mandatory features: spatial multiplexing, frame aggregation and block acknowledgement that are designed to improve the efficiency and throughput. We review these features, calculate the theoretical performance limits and present a comparative analysis of the performance gains that these features bring. We also present insights gained from measurements using a real 802. 11 n network. Using VolP as an example, the net effect on the application layer performance is evaluated.

Keywords: 802.11n, MCS, 802.11n Throughput, Frame Aggregation, BlockAcknowledgement, VoIP.

Knowledge and Experience Management in Action: A Case Study for Formulating User Resistance

Nasim Ghanbartehrani*, Venus Hosseinali-Mirza and Elahe Maleki
Industrial Management Institute, Tehran, Iran


This article uses an experience database as a basis for creating an organizational memory and knowledge development in a public sector organization. Re-applying the experience - knowledge which is obtained by solving problems that organization already solved - has an effective role in facilitating organizational activities and improving the efficiency. By applying this concept in a practical case, researchers model the experiences regarding user resistance against information system implementation and presents a framework for experience management - includes extract, save, utilize and update experiences. As the case study is a public sector agency, this framework could be used as a KM model for public sector organizations as well.

Keywords: Knowledge Management, Experience Management, IS Implementation, User Resistance, Experience Base.

Framework to Assess Customer Relationship Management Offering in the Application Service Provider Model

Eeshwari Bollapragada
Automatic Data Processing Inc., India.


The purpose of this paper is to build a framework that can independently be used by small and medium sized enterprises (SME) to assess application service providers (ASP) in the customer relationship management (CRM) space. The paper has primarily used the backdrop of the small and medium sized enterprises in the retail industry of India and has been able to build a framework for this segment. The paper lists lead management, contact management and key account management as the major sales functions that make or break a CRM purchase decision in an SME in the prevalent retail industry of India. Also, in the user interface perspective, security and ease of integration with legacy systems are the deciding factors for the adoption. The paper builds a framework, "AssessCRM" that any SME can use to assess their pain points and shortlist an ASP vendor based on the major considerations for their organization.

Keywords: Customer relationship management, application service providers, ASP, CRM, SME