Abstract: Online marketplaces are not only digital places where consumers buy and sell merchandise, and they are also destinations for brands to connect with real consumers at the moment when customers are in the shopping mindset. For many marketplaces, brands have been important partners through advertising. There can be, however, a risk of advertising impacting a consumer’s shopping journey if it hurts the use experience or takes the user away from the site. Both could lead to the loss of transaction revenue for the marketplace. In this paper, we present user-based methods for cannibalization control by selectively turning off ads to users who are likely to be cannibalized by ads subject to business objectives. We present ways of measuring cannibalization of advertising in the context of an online marketplace and propose novel ways of measuring cannibalization through purchase propensity and uplift modeling. A/B testing has shown that our methods can significantly improve user purchase and engagement metrics while operating within business objectives. To our knowledge, this is the first paper that addresses cannibalization mitigation at the user-level in the context of advertising.
Abstract: Cardiologists perform cardiac auscultation to detect
abnormalities in heart sounds. Since accurate auscultation is
a crucial first step in screening patients with heart diseases,
there is a need to develop computer-aided detection/diagnosis
(CAD) systems to assist cardiologists in interpreting heart sounds
and provide second opinions. In this paper different algorithms
are implemented for automated heart sound classification using
unsegmented phonocardiogram (PCG) signals. Support vector
machine (SVM), artificial neural network (ANN) and cartesian
genetic programming evolved artificial neural network (CGPANN)
without the application of any segmentation algorithm has been
explored in this study. The signals are first pre-processed to remove
any unwanted frequencies. Both time and frequency domain features
are then extracted for training the different models. The different
algorithms are tested in multiple scenarios and their strengths and
weaknesses are discussed. Results indicate that SVM outperforms
the rest with an accuracy of 73.64%.
Abstract: This paper presents a method for improving object search accuracy using a deep learning model. A major limitation to provide accurate similarity with deep learning is the requirement of huge amount of data for training pairwise similarity scores (metrics), which is impractical to collect. Thus, similarity scores are usually trained with a relatively small dataset, which comes from a different domain, causing limited accuracy on measuring similarity. For this reason, this paper proposes a deep learning model that can be trained with a significantly small amount of data, a clustered data which of each cluster contains a set of visually similar images. In order to measure similarity distance with the proposed method, visual features of two images are extracted from intermediate layers of a convolutional neural network with various pooling methods, and the network is trained with pairwise similarity scores which is defined zero for images in identical cluster. The proposed method outperforms the state-of-the-art object similarity scoring techniques on evaluation for finding exact items. The proposed method achieves 86.5% of accuracy compared to the accuracy of the state-of-the-art technique, which is 59.9%. That is, an exact item can be found among four retrieved images with an accuracy of 86.5%, and the rest can possibly be similar products more than the accuracy. Therefore, the proposed method can greatly reduce the amount of training data with an order of magnitude as well as providing a reliable similarity metric.
Abstract: Datasets or collections are becoming important assets by themselves and now they can be accepted as a primary intellectual output of a research. The quality and usage of the datasets depend mainly on the context under which they have been collected, processed, analyzed, validated, and interpreted. This paper aims to present a collection of program educational objectives mapped to student’s outcomes collected from self-study reports prepared by 32 engineering programs accredited by ABET. The manual mapping (classification) of this data is a notoriously tedious, time consuming process. In addition, it requires experts in the area, which are mostly not available. It has been shown the operational settings under which the collection has been produced. The collection has been cleansed, preprocessed, some features have been selected and preliminary exploratory data analysis has been performed so as to illustrate the properties and usefulness of the collection. At the end, the collection has been benchmarked using nine of the most widely used supervised multiclass classification techniques (Binary Relevance, Label Powerset, Classifier Chains, Pruned Sets, Random k-label sets, Ensemble of Classifier Chains, Ensemble of Pruned Sets, Multi-Label k-Nearest Neighbors and Back-Propagation Multi-Label Learning). The techniques have been compared to each other using five well-known measurements (Accuracy, Hamming Loss, Micro-F, Macro-F, and Macro-F). The Ensemble of Classifier Chains and Ensemble of Pruned Sets have achieved encouraging performance compared to other experimented multi-label classification methods. The Classifier Chains method has shown the worst performance. To recap, the benchmark has achieved promising results by utilizing preliminary exploratory data analysis performed on the collection, proposing new trends for research and providing a baseline for future studies.
Abstract: Intrusion detection systems (IDS) are the main components of network security. These systems analyze the network events for intrusion detection. The design of an IDS is through the training of normal traffic data or attack. The methods of machine learning are the best ways to design IDSs. In the method presented in this article, the pruning algorithm of C5.0 decision tree is being used to reduce the features of traffic data used and training IDS by the least square vector algorithm (LS-SVM). Then, the remaining features are arranged according to the predictor importance criterion. The least important features are eliminated in the order. The remaining features of this stage, which have created the highest level of accuracy in LS-SVM, are selected as the final features. The features obtained, compared to other similar articles which have examined the selected features in the least squared support vector machine model, are better in the accuracy, true positive rate, and false positive. The results are tested by the UNSW-NB15 dataset.
Abstract: Electricity prices have sophisticated features such as
high volatility, nonlinearity and high frequency that make forecasting
quite difficult. Electricity price has a volatile and non-random
character so that, it is possible to identify the patterns based on the
historical data. Intelligent decision-making requires accurate price
forecasting for market traders, retailers, and generation companies.
So far, many shallow-ANN (artificial neural networks) models have
been published in the literature and showed adequate forecasting
results. During the last years, neural networks with many hidden
layers, which are referred to as DNN (deep neural networks) have
been using in the machine learning community. The goal of this
study is to investigate electricity price forecasting performance of the
shallow-ANN and DNN models for the Turkish day-ahead electricity
market. The forecasting accuracy of the models has been evaluated
with publicly available data from the Turkish day-ahead electricity
market. Both shallow-ANN and DNN approach would give successful
result in forecasting problems. Historical load, price and weather
temperature data are used as the input variables for the models.
The data set includes power consumption measurements gathered
between January 2016 and December 2017 with one-hour resolution.
In this regard, forecasting studies have been carried out comparatively
with shallow-ANN and DNN models for Turkish electricity markets
in the related time period. The main contribution of this study
is the investigation of different shallow-ANN and DNN models
in the field of electricity price forecast. All models are compared
regarding their MAE (Mean Absolute Error) and MSE (Mean Square)
results. DNN models give better forecasting performance compare to
shallow-ANN. Best five MAE results for DNN models are 0.346,
0.372, 0.392, 0,402 and 0.409.
Abstract: Green microgrids using mostly renewable energy (RE) for generation, are complex systems with inherent nonlinear dynamics. Among a variety of different optimization tools there are only a few ones that adequately consider this complexity. This paper evaluates applicability of two somewhat similar optimization tools tailored for standalone RE microgrids and also assesses a machine learning tool for performance prediction that can enhance the reliability of any chosen optimization tool. It shows that one of these microgrid optimization tools has certain advantages over another and presents a detailed routine of preparing input data to simulate RE microgrid behavior. The paper also shows how neural-network-based predictive modeling can be used to validate and forecast solar power generation based on weather time series data, which improves the overall quality of standalone RE microgrid analysis.
Abstract: Power distribution circuits undergo frequent network
topology changes that are often left undocumented. As a result, the
documentation of a circuit’s connectivity becomes inaccurate with
time. The lack of reliable circuit connectivity information is one of the
biggest obstacles to model, monitor, and control modern distribution
systems. To enhance the reliability and efficiency of electric power
distribution systems, the circuit’s connectivity information must be
updated periodically. This paper focuses on one critical component of
a distribution circuit’s topology - the secondary transformer to phase
association. This topology component describes the set of phase lines
that feed power to a given secondary transformer (and therefore a
given group of power consumers). Finding the documentation of this
component is call Phase Identification, and is typically performed
with physical measurements. These measurements can take time
lengths on the order of several months, but with supervised learning,
the time length can be reduced significantly. This paper compares
several such methods applied to Phase Identification for a large
range of real distribution circuits, describes a method of training
data selection, describes preprocessing steps unique to the Phase
Identification problem, and ultimately describes a method which
obtains high accuracy (> 96% in most cases, > 92% in the worst
case) using only 5% of the measurements typically used for Phase
Identification.
Abstract: Where human beings can easily learn and adopt pronunciation variations, machines need training before put into use. Also humans keep minimum vocabulary and their pronunciation variations are stored in front-end of their memory for ready reference, while machines keep the entire pronunciation dictionary for ready reference. Supervised methods are used for preparation of pronunciation dictionaries which take large amounts of manual effort, cost, time and are not suitable for real time use. This paper presents an unsupervised adaptation model for building agile and dynamic pronunciation dictionaries online. These methods mimic human approach in learning the new pronunciations in real time. A new algorithm for measuring sound distances called Dynamic Phone Warping is presented and tested. Performance of the system is measured using an adaptation model and the precision metrics is found to be better than 86 percent.
Abstract: Hand gesture recognition is a technique used to locate, detect, and recognize a hand gesture. Detection and recognition are concepts of Artificial Intelligence (AI). AI concepts are applicable in Human Computer Interaction (HCI), Expert systems (ES), etc. Hand gesture recognition can be used in sign language interpretation. Sign language is a visual communication tool. This tool is used mostly by deaf societies and those with speech disorder. Communication barriers exist when societies with speech disorder interact with others. This research aims to build a hand recognition system for Lesotho’s Sesotho and English language interpretation. The system will help to bridge the communication problems encountered by the mentioned societies. The system has various processing modules. The modules consist of a hand detection engine, image processing engine, feature extraction, and sign recognition. Detection is a process of identifying an object. The proposed system uses Canny pruning Haar and Haarcascade detection algorithms. Canny pruning implements the Canny edge detection. This is an optimal image processing algorithm. It is used to detect edges of an object. The system employs a skin detection algorithm. The skin detection performs background subtraction, computes the convex hull, and the centroid to assist in the detection process. Recognition is a process of gesture classification. Template matching classifies each hand gesture in real-time. The system was tested using various experiments. The results obtained show that time, distance, and light are factors that affect the rate of detection and ultimately recognition. Detection rate is directly proportional to the distance of the hand from the camera. Different lighting conditions were considered. The more the light intensity, the faster the detection rate. Based on the results obtained from this research, the applied methodologies are efficient and provide a plausible solution towards a light-weight, inexpensive system which can be used for sign language interpretation.
Abstract: Image retrieval is the most interesting technique which is being used today in our digital world. CBIR, commonly expanded as Content Based Image Retrieval is an image processing technique which identifies the relevant images and retrieves them based on the patterns that are extracted from the digital images. In this paper, two research works have been presented using CBIR. The first work provides an automated and interactive approach to the analysis of CBIR techniques. CBIR works on the principle of supervised machine learning which involves feature selection followed by training and testing phase applied on a classifier in order to perform prediction. By using feature extraction, the image transforms such as Contourlet, Ridgelet and Shearlet could be utilized to retrieve the texture features from the images. The features extracted are used to train and build a classifier using the classification algorithms such as Naïve Bayes, K-Nearest Neighbour and Multi-class Support Vector Machine. Further the testing phase involves prediction which predicts the new input image using the trained classifier and label them from one of the four classes namely 1- Normal brain, 2- Benign tumour, 3- Malignant tumour and 4- Severe tumour. The second research work includes developing a tool which is used for tumour stage identification using the best feature extraction and classifier identified from the first work. Finally, the tool will be used to predict tumour stage and provide suggestions based on the stage of tumour identified by the system. This paper presents these two approaches which is a contribution to the medical field for giving better retrieval performance and for tumour stages identification.
Abstract: Crude oil market is an immensely complex and dynamic environment and thus the task of predicting changes in such an environment becomes challenging with regards to its accuracy. A number of approaches have been adopted to take on that challenge and machine learning has been at the core in many of them. There are plenty of examples of algorithms based on machine learning yielding satisfactory results for such type of prediction. In this paper, we have tried to predict crude oil prices using Long Short-Term Memory (LSTM) based recurrent neural networks. We have tried to experiment with different types of models using different epochs, lookbacks and other tuning methods. The results obtained are promising and presented a reasonably accurate prediction for the price of crude oil in near future.
Abstract: Crop yield prediction is a paramount issue in
agriculture. The main idea of this paper is to find out efficient
way to predict the yield of corn based meteorological records.
The prediction models used in this paper can be classified into
model-driven approaches and data-driven approaches, according to
the different modeling methodologies. The model-driven approaches are based on crop mechanistic
modeling. They describe crop growth in interaction with their
environment as dynamical systems. But the calibration process of
the dynamic system comes up with much difficulty, because it
turns out to be a multidimensional non-convex optimization problem.
An original contribution of this paper is to propose a statistical
methodology, Multi-Scenarios Parameters Estimation (MSPE), for the
parametrization of potentially complex mechanistic models from a
new type of datasets (climatic data, final yield in many situations).
It is tested with CORNFLO, a crop model for maize growth. On the other hand, the data-driven approach for yield prediction
is free of the complex biophysical process. But it has some strict
requirements about the dataset.
A second contribution of the paper is the comparison of these
model-driven methods with classical data-driven methods. For this
purpose, we consider two classes of regression methods, methods
derived from linear regression (Ridge and Lasso Regression, Principal
Components Regression or Partial Least Squares Regression) and
machine learning methods (Random Forest, k-Nearest Neighbor,
Artificial Neural Network and SVM regression).
The dataset consists of 720 records of corn yield at county scale
provided by the United States Department of Agriculture (USDA) and
the associated climatic data. A 5-folds cross-validation process and
two accuracy metrics: root mean square error of prediction(RMSEP),
mean absolute error of prediction(MAEP) were used to evaluate the
crop prediction capacity.
The results show that among the data-driven approaches, Random
Forest is the most robust and generally achieves the best prediction
error (MAEP 4.27%). It also outperforms our model-driven approach
(MAEP 6.11%). However, the method to calibrate the mechanistic
model from dataset easy to access offers several side-perspectives.
The mechanistic model can potentially help to underline the stresses
suffered by the crop or to identify the biological parameters of interest
for breeding purposes. For this reason, an interesting perspective is
to combine these two types of approaches.
Abstract: The world-wide population of people over 60 years
of age is growing rapidly. The explosion is placing increasingly
onerous demands on individual families, multiple industries and
entire countries. Current, human-intensive approaches to eldercare
are not sustainable, but IoT and AI technologies can help. The
Knowledge Reactor (KR) is a contextual, data fusion engine built to
address this and other similar problems. It fuses and centralizes IoT
and System of Record/Engagement data into a reactive knowledge
graph. Cognitive applications and services are constructed with its
multiagent architecture. The KR can scale-up and scaledown, because
it exploits container-based, horizontally scalable services for graph
store (JanusGraph) and pub-sub (Kafka) technologies. While the KR
can be applied to many domains that require IoT and AI technologies,
this paper describes how the KR specifically supports the challenging
domain of cognitive eldercare. Rule- and machine learning-based
analytics infer activities of daily living from IoT sensor readings. KR
scalability, adaptability, flexibility and usability are demonstrated.
Abstract: In terms of ITS, information on link characteristic is an essential factor for plan or operation. But in practical cases, not every link has installed sensors on it. The link that does not have data on it is called “Missing Link”. The purpose of this study is to impute data of these missing links. To get these data, this study applies the machine learning method. With the machine learning process, especially for the deep learning process, missing link data can be estimated from present link data. For deep learning process, this study uses “Recurrent Neural Network” to take time-series data of road. As input data, Dedicated Short-range Communications (DSRC) data of Dalgubul-daero of Daegu Metropolitan Area had been fed into the learning process. Neural Network structure has 17 links with present data as input, 2 hidden layers, for 1 missing link data. As a result, forecasted data of target link show about 94% of accuracy compared with actual data.
Abstract: With recent trends in Big Data and advancements
in Information and Communication Technologies, the healthcare
industry is at the stage of its transition from clinician oriented to
technology oriented. Many people around the world die of cancer
because the diagnosis of disease was not done at an early stage.
Nowadays, the computational methods in the form of Machine
Learning (ML) are used to develop automated decision support
systems that can diagnose cancer with high confidence in a timely
manner. This paper aims to carry out the comparative evaluation
of a selected set of ML classifiers on two existing datasets: breast
cancer and cervical cancer. The ML classifiers compared in this study
are Decision Tree (DT), Support Vector Machine (SVM), k-Nearest
Neighbor (k-NN), Logistic Regression, Ensemble (Bagged Tree) and
Artificial Neural Networks (ANN). The evaluation is carried out based
on standard evaluation metrics Precision (P), Recall (R), F1-score and
Accuracy. The experimental results based on the evaluation metrics
show that ANN showed the highest-level accuracy (99.4%) when
tested with breast cancer dataset. On the other hand, when these
ML classifiers are tested with the cervical cancer dataset, Ensemble
(Bagged Tree) technique gave better accuracy (93.1%) in comparison
to other classifiers.
Abstract: Data mining has, over recent years, seen big advances because of the spread of internet, which generates everyday a tremendous volume of data, and also the immense advances in technologies which facilitate the analysis of these data. In particular, classification techniques are a subdomain of Data Mining which determines in which group each data instance is related within a given dataset. It is used to classify data into different classes according to desired criteria. Generally, a classification technique is either statistical or machine learning. Each type of these techniques has its own limits. Nowadays, current data are becoming increasingly heterogeneous; consequently, current classification techniques are encountering many difficulties. This paper defines new measure functions to quantify the resemblance between instances and then combines them in a new approach which is different from actual algorithms by its reliability computations. Results of the proposed approach exceeded most common classification techniques with an f-measure exceeding 97% on the IRIS Dataset.
Abstract: Bioassay is the measurement of the potency of a chemical substance by its effect on a living animal or plant tissue. Bioassay data and chemical structures from pharmacokinetic and drug metabolism screening are mined from and housed in multiple databases. Bioassay prediction is calculated accordingly to determine further advancement. This paper proposes a four-step preprocessing of datasets for improving the bioassay predictions. The first step is instance selection in which dataset is categorized into training, testing, and validation sets. The second step is discretization that partitions the data in consideration of accuracy vs. precision. The third step is normalization where data are normalized between 0 and 1 for subsequent machine learning processing. The fourth step is feature selection where key chemical properties and attributes are generated. The streamlined results are then analyzed for the prediction of effectiveness by various machine learning algorithms including Pipeline Pilot, R, Weka, and Excel. Experiments and evaluations reveal the effectiveness of various combination of preprocessing steps and machine learning algorithms in more consistent and accurate prediction.
Abstract: Many supervised machine learning tasks require
decision making across numerous different classes. Multi-class
classification has several applications, such as face recognition, text
recognition and medical diagnostics. The objective of this article is
to analyze an adapted method of Stacking in multi-class problems,
which combines ensembles within the ensemble itself. For this
purpose, a training similar to Stacking was used, but with three
levels, where the final decision-maker (level 2) performs its training
by combining outputs from the tree-based pair of meta-classifiers
(level 1) from Bayesian families. These are in turn trained by pairs
of base classifiers (level 0) of the same family. This strategy seeks to
promote diversity among the ensembles forming the meta-classifier
level 2. Three performance measures were used: (1) accuracy, (2)
area under the ROC curve, and (3) time for three factors: (a)
datasets, (b) experiments and (c) levels. To compare the factors,
ANOVA three-way test was executed for each performance measure,
considering 5 datasets by 25 experiments by 3 levels. A triple
interaction between factors was observed only in time. The accuracy
and area under the ROC curve presented similar results, showing
a double interaction between level and experiment, as well as for
the dataset factor. It was concluded that level 2 had an average
performance above the other levels and that the proposed method
is especially efficient for multi-class problems when compared to
binary problems.
Abstract: Software technology is developing rapidly which leads to the growth of various industries. Now-a-days, software-based applications have been adopted widely for business purposes. For any software industry, development of reliable software is becoming a challenging task because a faulty software module may be harmful for the growth of industry and business. Hence there is a need to develop techniques which can be used for early prediction of software defects. Due to complexities in manual prediction, automated software defect prediction techniques have been introduced. These techniques are based on the pattern learning from the previous software versions and finding the defects in the current version. These techniques have attracted researchers due to their significant impact on industrial growth by identifying the bugs in software. Based on this, several researches have been carried out but achieving desirable defect prediction performance is still a challenging task. To address this issue, here we present a machine learning based hybrid technique for software defect prediction. First of all, Genetic Algorithm (GA) is presented where an improved fitness function is used for better optimization of features in data sets. Later, these features are processed through Decision Tree (DT) classification model. Finally, an experimental study is presented where results from the proposed GA-DT based hybrid approach is compared with those from the DT classification technique. The results show that the proposed hybrid approach achieves better classification accuracy.