Adaptive Automated Teller Machines

Nowadays, Automated Teller Machines (ATMs) provide significant online support to bank customers. A limitation of ATM usage is that customers often have to wait in a queue, especially at ATMs installed at busy locations. Also, old people tend to consume more ATM usage time, possibly frustrating customers in the queue. In these situations, ATMs should “adapt” to the behavior of the customers to minimize the usage time. To this end, we apply data mining techniques to an ATM transaction dataset obtained from an international bank based in Kuwait. We pre-process this dataset, and convert it into a specific XML format to mine it through the ProM (process mining) tool. Our results reveal that customers withdraw money most frequently, followed by purchases (through an ATM card) and balance inquiry transactions. Customers re-do these transactions frequently, and also employ them one after the other. We acquire the distributions of the withdrawn amount, based on individual customers, the location (ATM terminal) and time of the withdrawl. Based on these results, we have proposed a set of five adaptive ATM interfaces, which show only frequent transactions and frequently-withdrawn amounts, display the current balance autonomously, and query explicitly for viewing purchase history, or for performing another withdrawl. An online survey on 216 ATM customers reveals that a majority of customers are willing to use these interfaces for minimizing their usage time. Our work has been approved by the banking authority of Pakistan, and we are currently implementing our interfaces for a Pakistani bank.     

Image-driven virtual simulation of arthroscopy

In recent years, minimally invasive arthroscopic surgery has replaced a number of conventional open orthopedic surgery procedures on joints. While this achieves a number of advantages for the patient, the surgeons have to learn very different skills, since the surgery is performed with special miniature pencil-like tools and cameras inserted through little incisions while observing the surgical field on video monitor. Therefore, virtual reality simulation becomes an alternative to traditional surgical training based on hundreds years old apprentice–master model that involves either real patients or increasingly difficult to procure cadavers. Normally, 3D simulation of the virtual surgical field requires significant efforts from the software developers but yet remains not always photorealistic. In contrast to this, for photorealistic visualization and haptic interaction with the surgical field we propose to use real arthroscopic images augmented with 3D object models. The proposed technique allows for feeling the joint cavity displayed on video monitor as real 3D objects rather than their images while various surgical procedures, such as menisectomy, are simulated in real time. In the preprocessing stage of the proposed approach, the arthroscopic images are stitched into panoramas and augmented with implicitly defined object models representing deformable menisci. In the simulation loop, depth information from the mixed scene is used for haptic rendering. The scene depth map and visual display are reevaluated only when the scene is modified.     

The mask of (2b + 4)-point n-ary subdivision scheme

In this paper, we present general formulae for the mask of (2b + 4)-point n-ary approximating as well as interpolating subdivision schemes for any integers ${b\,\geqslant\,0}$ and ${n\,\geqslant\,2}$ . These formulae corresponding to the mask not only generalize and unify several well-known schemes but also provide the mask of higher arity schemes. Moreover, the 4-point and 6-point a-ary schemes introduced by Lian [Appl Appl Math Int J 3(1):18–29, 2008] are special cases of our general formulae.     

Short-term and long-term memory analysis of learning using 2D and 3D educational contents

ABSTRACT

The effect of 2D and 3D educational content learning on memory has been studied using electroencephalography (EEG) brain signal. A hypothesis is set that the 3D materials are better than the 2D materials for learning and memory recall. To test the hypothesis, we proposed a classification system that will predict true or false recall for short-term memory (STM) and long-term memory (LTM) after learning by either 2D or 3D educational contents. For this purpose, EEG brain signals are recorded during learning and testing; the signals are then analysed in the time domain using different types of features in various frequency bands. The features are then fed into a support vector machine (SVM)-based classifier. The experimental results indicate that the learning and memory recall using 2D and 3D contents do not have significant differences for both the STM and the LTM.     

Defect Prediction Using Akaike and Bayesian Information Criterion

Data available in software engineering for many applications contains variability and it is not possible to say which variable helps in the process of the prediction. Most of the work present in software defect prediction is focused on the selection of best prediction techniques. For this purpose, deep learning and ensemble models have shown promising results. In contrast, there are very few researches that deals with cleaning the training data and selection of best parameter values from the data. Sometimes data available for training the models have high variability and this variability may cause a decrease in model accuracy. To deal with this problem we used the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) for selection of the best variables to train the model. A simple ANN model with one input, one output and two hidden layers was used for the training instead of a very deep and complex model. AIC and BIC values are calculated and combination for minimum AIC and BIC values to be selected for the best model. At first, variables were narrowed down to a smaller number using correlation values. Then subsets for all the possible variable combinations were formed. In the end, an artificial neural network (ANN) model was trained for each subset and the best model was selected on the basis of the smallest AIC and BIC value. It was found that combination of only two variables’ ns and entropy are best for software defect prediction as it gives minimum AIC and BIC values. While, nm and npt is the worst combination and gives maximum AIC and BIC values.     

Hybrid model of content extraction

We present a hybrid model for content extraction from HTML documents. The model operates on Document Object Model (DOM) tree of the corresponding HTML document. It evaluates each tree node and associated statistical features like link density and text distribution across the node to predict significance of the node towards overall content provided by the document. Once significance of the nodes is determined, the formatting characteristics like fonts, styles and the position of the nodes are evaluated to identify the nodes with similar formatting as compared to the significant nodes. The proposed hybrid model is derived from two different models, i.e., one is based on statistical features and other on formatting characteristics and achieved the best accuracy. We describe the validity of model with the help of experiments conducted on the standard data sets. The results revealed that the proposed model outperformed other existing content extraction models. We present a browser based implementation of the proposed model as proof of concept and compare the implementation strategy with various state of art implementations. We also discuss various applications of the proposed model with special emphasis on open source intelligence.     

Study on pharyngeal and uvular consonants in foreign accented Arabic for ASR

This paper investigates the unique pharyngeal and uvular consonants of Arabic from the point of view of automatic speech recognition (ASR). Comparisons of the recognition error rates for these phonemes are analyzed in five experiments that involve different combinations of native and non-native Arabic speakers. The most three confusing consonants for every investigated consonant are discussed. All experiments use the Hidden Markov Model Toolkit (HTK) and the Language Data Consortium (LDC) WestPoint Modern Standard Arabic (MSA) database. Results confirm that these Arabic distinct consonants are a major source of difficulty for Arabic ASR. While the recognition rate for certain of these unique consonants such as // can drop below 35% when uttered by non-native speakers, there is advantage to include non-native speakers in ASR. Besides, regional differences in pronunciation of MSA by native Arabic speakers require the attention of Arabic ASR research.     

Rice-like CuO nanostructures for sensitive electrochemical sensing of hydrazine

In this work, a low temperature aqueous chemical growth methodology was used for the fabrication of CuO nanostructures. The as-synthesised nanostructures were then elaborately characterised by number of analytical techniques such as scanning electron microscopy (SEM) and X-ray powder diffraction (XRD). The obtained nanostructures were observed to possess interlaced rice-shaped structural features with the length and width of individual rice determined to be in the range of 200–300 nm and 50–100 nm respectively. The unique nanostructures when utilised as electrode material exhibited excellent electro-catalytic potential towards oxidation of hydrazine in alkaline media. The excellent conductive of CuO added by the high surface area of obtained nanorice-like structures enabled development of highly sensitive (3087 µA mM⁻¹ cm⁻²), selective and stable electrochemical sensor for hydrazine. In addition, the successfully application of the developed sensor in spiked tap, bottled and industrial water samples for the detection of hydrazine suggested its feasibility for practical environmental application.