Electrical Engineering

Each mobile device represents the digital footprint of the owner; at the same time mobile location data stored in telecom operators' databases in terms of Call Detail Record (CDR). It holds the precise identity of the mobile cell tower to which the owner is connected at any given time. Effectively mobile device count within a region for some time period can be calculated. Again, International Mobile Equipment Identity (IMEI) number is an unique identity to every mobile device, a part of which, known as Type Allocation Code (TAC) uniquely identifies the make and model of the mobile device which further identify the company or manufacturer of the mobile device. So combining them it is possible to analyze different business information about mobile penetration of companies in a defined region; hence the localized market share comparisons with other companies as well as among different models of same company. In order to model the problem and analyze huge CDR data, an analytical processing is carried out here using data warehouse. Here we propose a suitable Data warehouse schema which comprises of the required dimensions along with their concept hierarchies. The ETL processing which is done to form the data warehouse is described here. Finally the lattice of cuboids is constructed to carry out the OLAP processing from all possible business perspective.

Retaining an existing customer than acquiring a new one is less expensive in terms of marketing cost, bonus, and incentives offered etc. Telecommunication industry across the globe is stiff competitive environment where market is almost saturated and main focus of customer service becomes retaining existing customers and snatching others' customers to increase market share as well as profit. At the same time telecom industry facing the problem of churn or attrition more than anything especially for prepaid customers as the customers could switch the service providers easily. Telecom operators generate huge volume of call detail records every day for every call, SMS or internet access made by its customers. Telecom operators use these huge operational data for business processing to understand customer behavior. In this paper a mechanism is developed to store CDR data in a suitable Data Warehouse (DW) schema and analytically process these using OLAP tools to understand the prepaid customers usage, spending and propensity to marketing offers. Depending on the usage pattern proper segmentation of the customers has been done and they have been categorized for different types of targeted marketing offers and benefits to retain as well as acquire new customers.

This paper analyses the correlation between two different sectoral indices (e.g. between Automobile sector index and between Metal sector index, between Bank sector index and IT sectoral index etc.) in a time lagged manner. Lagging period is varied from 1 day to 5 days to investigate if any selected sector has lagged influence over any other sectoral index movement. If any up-ward/downward movement of a sectoral index (sector A) is correlated with sim-ilar upward/downward movement of another sectoral index (Sector B) with a time lag of ‘d’ days, then with association rule mining support and confidence is calculated for the combination. If d is the lag for which support and confidence is maximum then depending on the higher correlation as well as higher support and confidence value it is possible to forecast future (d days ahead of current day) movement of sector B based on present day movement of sector A. This model first uses correlational analysis to identify the level of dependence among two different sectors, then considers only those sectors having higher value of correlation for association rule mining. Those sector are not considered for which combination correlation is very low or 0.
This model has been tested with Indian share market data (NSE sectoral in-dex data of 6 sectors) of 2015. Result shows it is possible to predict in short term (1 to 5 days in future) price movement of sectoral indices using other lagged correlated sector price index movement.

A new steganographic model using dual image has been proposed to increase the obscurity of the secret as well as to make it more secure. It uses two different images namely reference image and cover image along with a secret embed key (stego key). Reference image is divided into blocks with assigned block-codes. Total number of blocks and other embedding parameters (block traversing direction, starting block etc.) are stored in the embed key. Secret message converted to binary and bit pairs are made. Bit pairs of secret message are encoded using the bit pairs of different blocks of reference image and sometimes updating a few LSB bits in the reference image. As an optional capacity enhancement module for text only secret messages dictionary word indexing based encoding is applied. Finally encoded bit stream is embedded into the cover image using any standard LSB technique with its own merits and demerits. LSB changes in the reference image are almost negligible. Embedding key is encrypted using public key of the receiver hence even if someone gets the encoded message it is not possible to get the secret message. As this model can use any existing LSB replacement scheme to embed encoded message into the cover image the state of the art security and other quality features are by default available to it. This is made more secure by using different dictionary word indexing scheme every time based on a random secret key that is appended in the embed key. Keywords—dual image; lsb steganography; reference image; dictionary word indexing; block code based encoding

Point of Sales (POS) terminals are provided by the banking and financial institutes to perform cashless transactions. Over the time due to different conveniences , use of digital money and online card transactions increased many folds. After each successful payment transaction at the POS terminals, a transaction log is sent to the POS terminal provider with payment-related financial data such as date and time, amount, authorization service provider, cardholder's bank, merchant identifier, store identifier, terminal number, etc. These data are useful for analytical processing which are useful for business. This paper proposes to process these huge transactional data using ETL process and thereafter construction of a data warehouse (DW) which enables the POS provider to employ certain analytical processing for business gain such as knowing own market share as well as position in market with respect to card payments, geographic location-wise business profiling, own as well as competitor's customer segmentation based on monthly card usage, monthly amount spent, etc.

Blood donation is an important issue throughout the world to save and manage lives. In order to make this process more effective an automated system could be built to monitor and organize blood donation camps. In this paper we developed an integrated framework with all related but isolated web based sub systems of a blood management system. We propose a data warehouse (DW) as an integral part of the integrated framework to store historical blood donation data in a centralized database for analytical processing. Proposed system would enable the authorities to take informed blood donation camping decision based on the analytical reports from the DW for some area for a particular time and citizen demography. Finally, we introduce a new measure of humanity (scoring system for good deeds) of citizens called Philanthropy Score (PS) and Philanthropy League (PL) derived from PS. In our proposed system after blood collection at donation camps, few health related details would be updated in the blood management system database and most importantly PS of the donor would be updated into the national citizens' database. A well-advertised communication would allow the citizens to know about the PS points they can accrue for every possible good deed. Blood donation is merely a prototype use of the proposed PS.

A very important issue with the e-commerce delivery service in most of the emerging economies including India is the last mile connectivity. Delivering products, booked online to the remote tier-2 and tier-3 cities remained "costly". It is observed from firsthand experience with some well-known e-commerce brands in India that their delivery service partners tend to cancel orders that are far away from their tier-2 logistics hubs with the reason shown as "address out of delivery range". Due to low order density in the far flanges of tier-2 and tier-3 cities arranging vehicles and delivery personnel become costly. In this paper, we propose an innovative delivery model to serve the remote areas by opening edge-hubs at selected places and employing local daily commuters for last mile delivery. Identifying the edge-hubs for opening distribution centers is a costly business if done using traditional field surveys. Here we propose the use of telecom call detail record (CDR) location data as an alternate way of identifying the hubs in real time with much less cost and time.

The objective of this paper is identifying a warehouse model to build an analytical framework and analyze different important parameters which directly impact the changes of share market. We identify parameters that represent different viewing windows and perspectives towards stock market performance and movement trends. We categorize and define many intrinsic as well as external factors that may affect stock market as a whole. Sensex and Nifty are used as the pulse of Indian stock market. In this paper, we focus on defining a suitable OLAP model which can cater all the parameters that affect share market. We also identify different applications of this analytical
model for forecasting information to help decision making.

A new steganography scheme using a reference image is proposed with two variants. In the first kind, secret text bits are directly embedded into the cover image after encryption using a reference image. In the second variant secret text is written/printed on a white canvas and then that image containing the secret text is converted into binary (BW) image. Any image could be used as a cover image. A special reference image will act as an encryption key and it is assumed that the used reference image is known to sender and receiver apriori. For a grayscale reference image, any k LSB bit planes could be used to encode the secret binary image bits before embedding them into k LSBs of the cover image. When k=1, the dimension of the secret image becomes the same as that of the reference image and cover image. With k=2, secret binary image size could be doubled or two secret images could be hidden using two LSB bit planes of the reference image and embedding them into two LSB bits of each cover image pixel. Any reversible mechanism could be used for encoding. Here XOR operation will be used for experiments. Two distinct advantages of this generic scheme are (i) It is far more secure than other image steganography techniques as adversary may know the encoding formulae , may collect the stego image but could not get information about the reference image used and which and how many bit planes are used for encoding as there could be millions of images and many combinations of bit planes that could have been employed; (ii) As the secret text is printed and then converted into an image, it overcomes one of the main limitations of LSB image steganography, i.e. robustness to random noise. Even with corrupted Stego image, the extracted binary image becomes corrupted too but most of the secret are still readable.
This scheme has lower capacity than the capacity that would have been achieved by directly embedding the secret text binary as in the first variant, but this could be overcome by using k>=2 as k=2 doubles the capacity. This whole scheme is generic and any digital media could be used as reference or cover. Experimental results demonstrate the robustness of the second variant against random noise, and standard image quality metrics such as MSE, PSNR, SSIM are evaluated and compared between the two variants.

Pixel value differencing (PVD) and Least Significant Bit (LSB) embedding are well known spatial domain stegano-graphic techniques. PVD utilizes the sharp changes of intensities among adjacent pixels where a large number of secret bits could be embedded without any perceptible change. one-time pad (OTP) symmetric cryptography is known for its security. In the proposed scheme one or more LSB bits of the selected pixels are used depending on the pixel intensity difference with neighboring pixels in 2 × 2 image blocks of the cover image. Secret bits are encrypted using OTP with randomly generated pre-shared key. Such encrypted bits are completely random and resemble noise hence make the scheme robust against different statistical attacks. Comparative simulations with some well-known PVD-based techniques show good results in terms of visual imperceptibility and different quality metrics such as MSE, PSNR, SSIM etc.

Stock market prediction is one of the most difficult computations due to the many internal as well as any number and type of external factors. It is impossible to get the exact computation hence we look for the  method which gives the computation with less error. Different machine
learning methods are being applied for the computations which involve many parameters. In this research work we choose Long Short-Term Memory (LSTM) for the prediction as it is computationally suitable for these types of data analysis. After doing the prediction of share price the work is extended to manage portfolio of the mutual fund. The framework has been designed in such a way so that the portfolio manager can choose any number of business sectors as well as any number of shares
belong to this sector. This research work henceforth applicable for computing individual share price as well as managing a diversified portfolio.

Investment in the share market helps generate more profit than the other financial instruments but has the threat of market risk that might lead to a high loss. This risk factor refrains many potential investors from investing in the share market directly. Instead, they invest in different mutual funds that are being managed by experienced portfolio managers. To avoid the risk factors and increase the gain, they put the accumulated capital in multiple stocks. They need to perform many calculations and predictions to overcome the uncertainties and unpredictability and need to ensure higher gains to the investors of that mutual fund. In this research work initially, a data mining based approach employs a curve fitting/regression technique to forecast the individual stock price. Based on the above analysis, we propose a framework to diversify the investment of the capital fund. This method employs buy and hold strategy using both statistical features and basic domain knowledge of the share market. The proposed framework distributes the capital first, by distributing sector-wise, and then for each sector, investing company-wise, as a diversified approach among different stocks for higher return but maintaining lower risks. Experimental results show that the proposed framework performs well and generates a good yield compared to some benchmark and ranked mutual funds in the Indian stock market.

In this new era of Web 2.0, people around the world are expressing their feelings, sentiments, thoughts, daily activities, and local and global events happening around them in different social networking sites like Twitter, Facebook, etc. This generates vast amount of data in social media by registered users which are geographical and temporal information-oriented. This rich data could be potentially useful information and is being extensively used nowadays for different applications like user’s sentiment analysis, product or service reviews, real-time information extraction like traffic, disaster reporting, personalized message or user recommendation, and other areas. Extracting topic distribution from social media in spatial and temporal dimensions is an important research area. Hence, our focus of this study is on discussing various topical modeling techniques and their uses in different recent research works. This chapter gives a brief overview of the recent updates of spatiotem...

This paper analyses the correlation between two different sectoral indices (e.g. between Automobile sector index and between Metal sector index, between Bank sector index and IT sectoral index etc.) in a time lagged manner. Lagging period is varied from 1 day to 5 days to investigate if any selected sector has lagged influence over any other sectoral index movement. If any upward/downward movement of a sectoral index (sector A) is correlated with similar upward/downward movement of another sectoral index (Sector B) with a time lag of ‗d' days, then with association rule mining support and confidence is calculated for the combination. If d is the lag for which support and confidence is maximum then depending on the higher correlation as well as higher support and confidence value it is possible to forecast future (d days ahead of current day) movement of sector B based on present day movement of sector A. This model first uses correlational analysis to identify the level of dependence among two different sectors, then considers only those sectors having higher value of correlation for association rule mining. Those sector are not considered for which combination correlation is very low or 0. This model has been tested with Indian share market data (NSE sectoral index data of 6 sectors) of 2015. Result shows it is possible to predict in short term (1 to 5 days in future) price movement of sectoral indices using other lagged correlated sector price index movement.

A very important issue with the e-commerce delivery service in most of the emerging economies including India is the last mile connectivity. Delivering products, booked online to the remote tier-2 and tier-3 cities remained “costly”. It is observed from firsthand experience with some well-known e-commerce brands in India that their delivery service partners tend to cancel orders that are far away from their tier-2 logistics hubs with the reason shown as “address out of delivery range”. Due to low order density in the far flanges of tier-2 and tier-3 cities arranging vehicles and delivery personnel become costly. In this paper, we propose an innovative delivery model to serve the remote areas by opening edge-hubs at selected places and employing local daily commuters for last mile delivery. Identifying the edge-hubs for opening distribution centers is a costly business if done using traditional field surveys. Here we propose the use of telecom call detail record (CDR) location data a...

Mining operational telecom data has given many interesting insights about customer behavior, mobility, social interaction etc and has great business value. But to have truly unbiased analysis over a defined geographic region, we need to first consolidate all telecom data from many different operators present within the region and then analyze them. This necessitates an integration architecture that will allows to integrate, consolidate and store different type of operational telecom data such as call detail record (CDR), customer relationship management (CRM) data, network data etc into a central data warehouse (DW). Though almost all operator stores semantically same type of data but in a little different form and format. Again these operators are separate business entities so integrating data from their operational databases and data marts needs to be independent of each other and flexible in fetching data for extract, transform and loading processes. In this paper we shall review a few existing DW schemas used to consolidate CDR and CRM data and propose a generic model DW schema with indentified facts and dimensions for the central DW. Next we propose integration architecture to collect data and consolidate in the central DW from many different operational databases of different telecom operators. Web services (WS) has been proposed as a mode of communication between central DW and other operational databases.

Identifying the influential spreaders in complex networks has
emerged as an important research challenge to control the spread of (mis)information or infectious diseases. Researchers have proposed many centrality measures to identify the influential nodes (spreaders), in the past few years. Still, most of them have not considered the importance of the edges in unweighted networks. To address this issue, we propose a novel centrality measure to identify the spreading ability of the Influential Spreaders using Potential Edge Weight (IS-PEW). Considering the connectivity structure, ability of information exchange, and the importance of neighboring nodes, we measure the potential edge weight. The ranking similarity of spreaders identified by the IS-PEW method and the baseline centrality methods is compared with Susceptible–Infectious–Recovered (SIR) epidemic simulator using Kendall’s rank correlation. Considering six different real networks, the spreading ability of the top-ranking spreaders is also compared for five different percentages of top-ranking node sets.