Search

In both public and private sectors, critical technology-based tasks, such as innovation, forecasting, and road-mapping, are faced with unmanageable complexity due to the ever-expanding web of technologies which can range into thousands. This context cannot be easily handled manually or with efficient speed. However, more precise and insightful answers are expected. These answers are the fundamental challenge addressed by tech-mining. For tech-mining, discovering the associations among them is a critical task. These associations are used to form a weighted directed graph of technologies called “association tech-graph” which is used for technology development, trend analysis, policymaking, strategic planning, and innovation. In this article, we present a novel method to build an artificial intelligence (AI) agent for automatic association discovery among technologies in a way that matches the quality of the human experts. To this end, neural network-based word embedding methods are exploited to represent technology terms as vectors, and their associations are calculated using similarity measures. To increase the accuracy of the vectors, several crawlers are built to acquire more appropriate training data. Furthermore, we introduce a validation method to measure the accuracy of the AI agent compared to human intelligence, which allows us to discuss the drawbacks of both approaches.

Many studies in the rigid gas permeable (RGP) lens fitting field have focused on providing the best fit for patients with irregular astigmatism, a challenging issue. Despite the ease and accuracy of fitting in the current fitting methods, no studies have provided a high-pace solution with the final best fit to assist experts. This work presents a deep learning solution for identifying features in Pentacam four refractive maps and RGP base curve identification. An authentic dataset of 247 samples of Pentacam four refractive maps was gathered, providing a multi-view image of the corneal structure. Scratch-based convolutional neural network (CNN) architectures and well-known CNN architectures such as AlexNet, GoogLeNet, and ResNet have been used to extract features and transfer learning. Features are aggregated through a fusion technique. Based on a comparison of means square error (MSE) of normalized labels, the multi-view scratch-based CNN provided R-squared of 0.849, 0.846, 0.835, and 0.834 followed by GoogLeNet, comparable with current methods. Transfer learning outperforms various scratch-based CNN models, through which proper specifications some scratch-based models were able to increase coefficient of determinations. CNNs on multi-view Pentacam images have enabled fast detection of the RGP lens base curve, higher patient satisfaction, and reduced chair time. Graphical abstract The Pentacam four refractive maps is learned by the proposed scratch-based and transfer learning-based CNN methodology. The deep network-based solutions enable identification of rigid gas permeable lens for patients with irregular astigmatism.

This research is aimed at accurate identification of base-curve in rigid gas-permeable (RGP) lens based on supervised image processing and classification of Pentacam four refractive maps in irregular astigmatism cases. Base-curve, is typically identified based on expert’s opinion of the corneal structure of the eye. Studies have applied time-consuming methods, focusing on manual and device-based techniques. For the identification of the base-curve of a lens, image analysis is proposed. As each map in the four refractive maps is of a singular view, multi-view learning is recommended to provide a single representation. To this end, an authentic dataset consisting of 247 labeled Pentacam four refractive maps was gathered in which labels were verified manually. We have proposed two novel feature extraction techniques in this domain: quantization-based radial–sectoral segmentation (QRSS) in image processing and deep convolutional neural networks. Feature fusion is applied and RGP base-curve is identified by the regression layer of a neural network. A combination of QRSS and multilayered perceptron delineates the best result, achieving a coefficient of determination of 0.9642 and satisfactory mean square error (0.0089) which is acceptable by the experts. The proposed multi-view model could improve base-curve detection accuracy, with less trial and error and patient visits in the lens fitting process.

Many traditional Botnet detection methods have trouble scaling up to meet the needs of multi-Gbps networks. This scalability challenge is not just limited to bottlenecks in the detection process, but across all individual components of the Botnet detection system including data gathering, storage, feature extraction, and analysis. In this paper, we propose a fully scalable big data framework that enables scaling for each individual component of Botnet detection. Our framework can be used with any Botnet detection method - including statistical methods, machine learning methods, and graph-based methods. Our experimental results show that the proposed framework successfully scales in live tests on a real network with 5Gbps of traffic throughput and 50 millions IP addresses visits. In addition, our run time scales logarithmically with respect to the volume of the input for example, when the scale of the input data multiplies by 4 × , the total run time increases by only 31%. This is significant improvement compared to schemes such as Botcluster in which run time increases by 86% under similar scale condition.

Today, the providers of cloud computing services are among the most prominent service suppliers worldwide. Availability of cloud services is one of the most important concerns of cloud service providers (CSPs) and cloud users (CUs). Distributed Denial of Service (DDoS) attacks are common types of security issues which affect cloud services and consequently, can lead to unavailability of the services. Therefore, reducing the effects of DDoS attacks helps CSPs to provide high quality services to CUs. In this paper, first, we propose an anomaly-based DDoS attack detection framework in cloud environment using a third party auditor (TPA). Second, we provide multiple basic assumptions and configurations of cloud environments for establishing simulation tests to evaluate our proposed framework. Then, we provide results of simulation tests to analyze the feasibility of our approach. Simulation results demonstrate that our method for detecting DDoS attacks in CSPs has following advantages: efficiency, because of the low overhead of computations on CSPs for attack detection; rapid, due to informing a CSP about an attack in a short course of time regarding the maximum valid response time which is defined in a service level agreement (SLA); and precision, through no false positive detection as well as a low rate of false negative detection which is

Purpose– Recommender systems are techniques that allow companies to develop sales and marketing and as a result, attract more customers. There are several different types of recommender systems which collaborative filtering (CF) method is more popular and is used in various fields. However, similar to other recommender systems, this system has its own limitations. Nowadays, recommender systems are combined with other systems to enhance the quality and precision. The purpose of this paper is to present a new method to increase the accuracy and quality of recommendations associated with filtering systems.Design/methodology/approach– First, the recency, frequency, and monetary (RFM) variables of the clients are extracted and variables’ weights are calculated. Then, using weighted RFM and expectation maximization clustering algorithms and their combination with the closest K-neighbors, recommendations for each cluster is independently extracted. Finally, the results are compared with the outcome of conventional CF techniques. Remarkably, sale transactions of a big distribution and sale of goods centers are used in this study.Findings– The results indicated that the proposed method has higher accuracy compared to the conventional CF method. Likewise, the clusters which have higher values were received more accurate recommendations. Another point was that the proposed method was faster on obtaining the results than the conventional method as the recommendations were performed with respect to the customers of the same cluster, while all clients were assessed in the conventional method and as a result, the calculation speed is reduced as the number of customers increases in this method.Originality/value– The results indicated that the proposed method has higher accuracy compared to the conventional CF method. Likewise, the clusters which have higher values were received more accurate recommendations. This is very important for businesses and trade centers as more than 80 percent of their profits come from valued customers and hence, recommendations with higher accuracy to these valued customers lead to more profits to sales centers. Since the valued customers were calculated in the proposed method and the value of each customer was distinguished for sales representatives, the accomplished recommendations can be coordinated with sales’ strategies to make it more targeted.

The RFM (Recency, Frequency and Monetary) model provides an effective analysis for decision makers in order to target their customers and develop appropriate marketing strategies according to their previous behaviors. Although the RFM model has been widely applied in various areas of marketing, its simplicity threatens its effectiveness since it does not consider the customers' relationship and changes in customers' behavior. In this paper, we propose an R+FM model which configures the segmentation according to the business changes and clusters customers using K-Means. We applied our model on Digikala company, the biggest E-Commerce in Middle East, and compared our model with the Digikala's previous RFM model which used Customer Quantile Method. Moreover, we built strategies for each segment and ran an SMS campaign according to those strategies. The results of the campaign showed that our Segmentation Model improved the number of purchase and average monetary of the baskets.

The popularity of smartphones around the world has the potential to dramatically improve healthcare, due to the high portability, computing capability, and ease of usage. Today's smartphones are easily programmable and come with a growing set of powerful embedded sensors such as accelerometers, gyroscopes, microphones, and cameras. Indeed, the smartphones equipped with such miniaturized sensors will potentially reshape the future of healthcare by facilitating proactive personal wellbeing management and ubiquitous health monitoring including physiological signs and human posture observation. In this paper, we present the design and implementation of a smartphone-centric software for monitoring the human posture by using the acceleration sensors which are embedded in smartphones. Additionally, an emphasis is given to interpreting the obtained data from the acceleration sensors to achieve context-awareness suitable for healthcare applications. Such the smartphone-centric monitoring softwares are also more cost-effective and less complex compared to its conventional counterparts where multiple wearable sensors are incorporated.

Torsion rods are a primary component of many power transmission and other mechanical systems. The behavior of these rods under elastoplastic torsion is of major concern for designers. Different methods have so far been proposed which deal with the elastoplastic torsion of rods, most of which assume constant yield stress. This assumption produces rough and inaccurate results when the rods are heat treated, since in the process of heat treatment the form of yield stress distribution across the rod cross section changes. We propose a new method for calculating elastoplastic torsion of rods of simply connected cross section which is based on heat treatment observations. In our method the full plastic stress function is obtained by using the semi-inverse method. Elastoplastic stress function is obtained by generalizing the idea of the membrane analogy and using a piecewise continuous stress function. Since the proposed form of yield stress distribution can not be handled by the current Finite Element packages, we produce a computer package with a 3D graphical interface capable of calculating and displaying the 3D elastoplastic stress function, shear stress contours, and torque-angle of rotation per unit length. We show that our method produces excellent agreement for several known cross sections in comparison to methods which assume constant yield stress.

We define localized content-based image retrieval as a CBIR task where the user is only interested in a portion of the image, and the rest of the image is irrelevant. In this paper we present a localized CBIR system, ACCIO, that uses labeled images in conjunction with a multiple-instance learning algorithm to first identify the desired object and weight the features accordingly, and then to rank images in the database using a similarity measure that is based upon only the relevant portions of the image. A challenge for localized CBIR is how to represent the image to capture the content. We present and compare two novel image representations, which extend traditional segmentation-based and salient point-based techniques respectively, to capture content in a localized CBIR setting.

We present Mi-Winnow, a new multiple-instance learning (MIL) algorithm that provides a new technique to convert MIL data into standard supervised data. In MIL each example is a collection (or bag) of d-dimensional points where each dimension corresponds to a feature. A label is provided for the bag, but not for the individual points within the bag. Mi-Winnow is different from existing multiple-instance learning algorithms in several key ways. First, Mi-Winnow allows each image to be converted into a bag in multiple ways to create training (and test) data that varies in both the number of dimensions per point, and in the kind of features used. Second, instead of learning a concept defined by a single point-and-scaling hypothesis, Mi-Winnow allows the underlying concept to be described by combining a set of separators learned by Winnow. For content-based image retrieval applications, such a generalized hypothesis is important since there may be different ways to recognize which images are of interest

Salient points are locations in an image where there is a significant variation with respect to a chosen image feature. Since the set of salient points in an image capture important local characteristics of that image, they can form the basis of a good image representation for content-based image retrieval (CBIR). The features for a salient point should represent the local characteristic of that point so that the similarity between features indicates the similarity between the salient points. Traditional uses of salient points for CBIR assign features to a salient point based on the image features of all pixels in a window around that point. However, since salient points are often on the boundary of objects, the features assigned to a salient point often involve pixels from different objects. In this paper, we propose a CBIR system that uses a novel salient point method that both reduces the number of salient points using a segmentation as a filter, and also improves the representation so that it is a more faithful representation of a single object (or portion of an object) that includes information about its surroundings. We also introduce an improved Expectation Maximization-Diverse Density (EM-DD) based multiple-instance learning algorithm. Experimental results show that our CBIR techniques improve retrieval performance by 5%-11% as compared with current methods.

There has been much work on applying multiple-instance (MI) learning to content-based image retrieval (CBIR) where the goal is to rank all images in a known repository using a small labeled data set. Most existing MI learning algorithms are non-transductive in that the images in the repository serve only as test data and are not used in the learning process. We present MISSL (Multiple-Instance Semi-Supervised Learning) that transforms any MI problem into an input for a graph-based single-instance semi-supervised learning method that encodes the MI aspects of the problem simultaneously working at both the bag and point levels. Unlike most prior MI learning algorithms, MISSL makes use of the unlabeled data.

Classic Content-Based Image Retrieval (CBIR) takes a single non-annotated query image, and retrieves similar images from an image repository. Such a search must rely upon a holistic (or global) view of the image. Yet often the desired content of an image is not holistic, but is localized. Specifically, we define Localized Content-Based Image Retrieval as a CBIR task where the user is only interested in a portion of the image, and the rest of the image is irrelevant. Many classic CBIR systems use relevance feedback to obtain images labeled as desirable or not desirable. Yet, these labeled images are typically used only to re-weight the features used within a global similarity measure. In this paper we present a localized CBIR system, acciop, that uses labeled images in conjunction with a multiple-instance learning algorithm to first identify the desired object and re-weight the features, and then to rank images in the database using a similarity measure that is based upon individual regions within the image. We evaluate our system using a five-category natural scenes image repository, and benchmark data set, SIVAL, that we have constructed with 25 object categories.