Search results for: categorical datasets

Authors: Semeh Ben Salem, Sami Naouali, Moetez Sallami

Abstract:

Clustering is a well known data mining technique used in pattern recognition and information retrieval. The initial dataset to be clustered can either contain categorical or numeric data. Each type of data has its own specific clustering algorithm. In this context, two algorithms are proposed: the k-means for clustering numeric datasets and the k-modes for categorical datasets. The main encountered problem in data mining applications is clustering categorical dataset so relevant in the datasets. One main issue to achieve the clustering process on categorical values is to transform the categorical attributes into numeric measures and directly apply the k-means algorithm instead the k-modes. In this paper, it is proposed to experiment an approach based on the previous issue by transforming the categorical values into numeric ones using the relative frequency of each modality in the attributes. The proposed approach is compared with a previously method based on transforming the categorical datasets into binary values. The scalability and accuracy of the two methods are experimented. The obtained results show that our proposed method outperforms the binary method in all cases.

Keywords: clustering, unsupervised learning, pattern recognition, categorical datasets, knowledge discovery, k-means

Procedia PDF Downloads 259

Authors: Murat Yazici

Abstract:

Anomalies are irregularities found in data that do not adhere to a well-defined standard of normal behavior. The identification of outliers or anomalies in data has been a subject of study within the statistics field since the 1800s. Over time, a variety of anomaly detection techniques have been developed in several research communities. The cluster analysis can be used to detect anomalies. It is the process of associating data with clusters that are as similar as possible while dissimilar clusters are associated with each other. Many of the traditional cluster algorithms have limitations in dealing with data sets containing categorical properties. To detect anomalies in categorical data, fuzzy clustering approach can be used with its advantages. The fuzzy k-Mode (FKM) clustering algorithm, which is one of the fuzzy clustering approaches, by extension to the k-means algorithm, is reported for clustering datasets with categorical values. It is a form of clustering: each point can be associated with more than one cluster. In this paper, anomaly detection is performed on two simulated data by using the FKM cluster algorithm. As a significance of the study, the FKM cluster algorithm allows to determine anomalies with their abnormality degree in contrast to numerous anomaly detection algorithms. According to the results, the FKM cluster algorithm illustrated good performance in the anomaly detection of data, including both one anomaly and more than one anomaly.

Keywords: fuzzy k-mode clustering, anomaly detection, noise, categorical data

Procedia PDF Downloads 53

Authors: Rishabh Srivastav, Divyam Sharma

Abstract:

We propose an algorithm to cluster categorical data named as ‘Genetic algorithm initialized rough set based fuzzy K-Modes for categorical data’. We propose an amalgamation of the simple K-modes algorithm, the Rough and Fuzzy set based K-modes and the Genetic Algorithm to form a new algorithm,which we hypothesise, will provide better Centroid Model clustering results, than existing standard algorithms. In the proposed algorithm, the initialization and updation of modes is done by the use of genetic algorithms while the membership values are calculated using the rough set and fuzzy logic.

Keywords: categorical data, fuzzy logic, genetic algorithm, K modes clustering, rough sets

Procedia PDF Downloads 246

Authors: Madaki Umar Yusuf, Mohd Rizam B. Abubakar

Abstract:

The survival censored data with incomplete covariate data is a common occurrence in many studies in which the outcome is survival time. With model when the missing covariates are categorical, a useful technique for obtaining parameter estimates is the EM by the method of weights. The survival outcome for the class of generalized linear model is applied and this method requires the estimation of the parameters of the distribution of the covariates. In this paper, we propose some clinical trials with ve covariates, four of which have some missing values which clearly show that they were fully censored data.

Keywords: EM algorithm, incomplete categorical covariates, ignorable missing data, missing at random (MAR), Weibull Distribution

Procedia PDF Downloads 405

Authors: M. A. Wanzala, L. A. Ogallo, F. J. Opijah, J. N. Mutemi

Abstract:

The usefulness and limitations in climate information are due to uncertainty inherent in the climate system. For any given region to have sustainable development it is important to apply climate information into its socio-economic strategic plans. The overall objective of the study was to assess the performance of the Coupled Model Inter-comparison Project (CMIP5) over the Lake Victoria Basin. The datasets used included the observed point station data, gridded rainfall data from Climate Research Unit (CRU) and hindcast data from eight CMIP5. The methodology included trend analysis, spatial analysis, correlation analysis, Principal Component Analysis (PCA) regression analysis, and categorical statistical skill score. Analysis of the trends in the observed rainfall records indicated an increase in rainfall variability both in space and time for all the seasons. The spatial patterns of the individual models output from the models of MPI, MIROC, EC-EARTH and CNRM were closest to the observed rainfall patterns.

Keywords: categorical statistics, coupled model inter-comparison project, principal component analysis, statistical downscaling

Procedia PDF Downloads 368

Authors: Doyin Afolabi, Phillip Adewole, Oladipupo Sennaike

Abstract:

Most well-known classifiers, including the decision tree algorithm, can make predictions on balanced datasets efficiently. However, the decision tree algorithm tends to be biased towards imbalanced datasets because of the skewness of the distribution of such datasets. To overcome this problem, this study proposes a weighted decision tree algorithm that aims to remove the bias toward the majority class and prevents the reduction of majority observations in imbalance datasets classification. The proposed weighted decision tree algorithm was tested on three imbalanced datasets- cancer dataset, german credit dataset, and banknote dataset. The specificity, sensitivity, and accuracy metrics were used to evaluate the performance of the proposed decision tree algorithm on the datasets. The evaluation results show that for some of the weights of our proposed decision tree, the specificity, sensitivity, and accuracy metrics gave better results compared to that of the ID3 decision tree and decision tree induced with minority entropy for all three datasets.

Keywords: data mining, decision tree, classification, imbalance dataset

Procedia PDF Downloads 136

Authors: Ghada A. Alfattni

Abstract:

Analysing unbalanced datasets is one of the challenges that practitioners in machine learning field face. However, many researches have been carried out to determine the effectiveness of the use of the synthetic minority over-sampling technique (SMOTE) to address this issue. The aim of this study was therefore to compare the effectiveness of the SMOTE over different models on unbalanced datasets. Three classification models (Logistic Regression, Support Vector Machine and Nearest Neighbour) were tested with multiple datasets, then the same datasets were oversampled by using SMOTE and applied again to the three models to compare the differences in the performances. Results of experiments show that the highest number of nearest neighbours gives lower values of error rates. 

Keywords: imbalanced datasets, SMOTE, machine learning, logistic regression, support vector machine, nearest neighbour

Procedia PDF Downloads 350

Authors: Saba Ebrahimi, Saeed Ahmadian, Hedie Ashrafi

Abstract:

Selecting appropriate patients for surgery is one of the main issues in thoracic surgery (TS). Both short-term and long-term risks and benefits of surgery must be considered in the patient selection criteria. There are some limitations in the existing datasets of TS patients because of missing values of attributes and imbalanced distribution of survival classes. In this study, a novel ensemble architecture of deep learning networks is proposed based on stacking different linear and non-linear layers to deal with imbalance datasets. The categorical and numerical features are split using different layers with ability to shrink the unnecessary features. Then, after extracting the insight from the raw features, a novel biased-kernel layer is applied to reinforce the gradient of the minority class and cause the network to be trained better comparing the current methods. Finally, the performance and advantages of our proposed model over the existing models are examined for predicting patient survival after thoracic surgery using a real-life clinical data for lung cancer patients.

Keywords: deep learning, ensemble models, imbalanced classification, lung cancer, TS patient selection

Procedia PDF Downloads 145

Authors: Mohd Asrul Affendi Bin Abdullah

Abstract:

Sample size calculation is especially important for zero inflated models because a large sample size is required to detect a significant effect with this model. This paper verify how to present percentage of power approximation for categorical and then extended to zero inflated models. Wald test was chosen to determine power sample size of AIDS death rate because it is frequently used due to its approachability and its natural for several major recent contribution in sample size calculation for this test. Power calculation can be conducted when covariates are used in the modeling ‘excessing zero’ data and assist categorical covariate. Analysis of AIDS death rate study is used for this paper. Aims of this study to determine the power of sample size (N = 945) categorical death rate based on parameter estimate in the simulation of the study.

Keywords: power sample size, Wald test, standardize rate, ZINBDR

Procedia PDF Downloads 435

Authors: Nushrat Khan, Mike Thelwall, Kayvan Kousha

Abstract:

Making research data openly accessible has been mandated by most funders over the last 5 years as it promotes reproducibility in science and reduces duplication of effort to collect the same data. There are evidence that articles that publicly share research data have higher citation rates in biological and social sciences. However, how and whether shared data is being reused is not always intuitive as such information is not easily accessible from the majority of research data repositories. This study aims to understand the practice of data citation and how data is being reused over the years focusing on biodiversity since research data is frequently reused in this field. Metadata of 38,878 datasets including citation counts were collected through the Global Biodiversity Information Facility (GBIF) API for this purpose. GBIF was used as a data source since it provides citation count for datasets, not a commonly available feature for most repositories. Analysis of dataset types, citation counts, creation and update time of datasets suggests that citation rate varies for different types of datasets, where occurrence datasets that have more granular information have higher citation rates than checklist and metadata-only datasets. Another finding is that biodiversity datasets on GBIF are frequently updated, which is unique to this field. Majority of the datasets from the earliest year of 2007 were updated after 11 years, with no dataset that was not updated since creation. For each year between 2007 and 2017, we compared the correlations between update time and citation rate of four different types of datasets. While recent datasets do not show any correlations, 3 to 4 years old datasets show weak correlation where datasets that were updated more recently received high citations. The results are suggestive that it takes several years to cumulate citations for research datasets. However, this investigation found that when searched on Google Scholar or Scopus databases for the same datasets, the number of citations is often not the same as GBIF. Hence future aim is to further explore the citation count system adopted by GBIF to evaluate its reliability and whether it can be applicable to other fields of studies as well.

Keywords: data citation, data reuse, research data sharing, webometrics

Procedia PDF Downloads 178

Authors: Michael Shulman, Paige North, Benedikt Ahrens, Dmitris Tsementzis

Abstract:

The Univalence Principle is the statement that equivalent mathematical structures are indistinguishable. We prove a general version of this principle that applies to all set-based, categorical, and higher-categorical structures defined in a non-algebraic and space-based style, as well as models of higher-order theories such as topological spaces. In particular, we formulate a general definition of indiscernibility for objects of any such structure, and a corresponding univalence condition that generalizes Rezk’s completeness condition for Segal spaces and ensures that all equivalences of structures are levelwise equivalences. Our work builds on Makkai’s First-Order Logic with Dependent Sorts, but is expressed in Voevodsky’s Univalent Foundations (UF), extending previous work on the Structure Identity Principle and univalent categories in UF. This enables indistinguishability to be expressed simply as identification, and yields a formal theory that is interpretable in classical homotopy theory, but also in other higher topos models. It follows that Univalent Foundations is a fully equivalence-invariant foundation for higher-categorical mathematics, as intended by Voevodsky.

Keywords: category theory, higher structures, inverse category, univalence

Procedia PDF Downloads 151

Authors: Najmeh Abedzadeh, Matthew Jacobs

Abstract:

An intrusion detection system (IDS) is a software application that monitors malicious activities and generates alerts if any are detected. However, most network activities in IDS datasets are normal, and the relatively few numbers of attacks make the available data imbalanced. Consequently, cyber-attacks can hide inside a large number of normal activities, and machine learning algorithms have difficulty learning and classifying the data correctly. In this paper, a comprehensive literature review is conducted on different types of algorithms for both implementing the IDS and methods in correcting the imbalanced IDS dataset. The most famous algorithms are machine learning (ML), deep learning (DL), synthetic minority over-sampling technique (SMOTE), and reinforcement learning (RL). Most of the research use the CSE-CIC-IDS2017, CSE-CIC-IDS2018, and NSL-KDD datasets for evaluating their algorithms.

Keywords: IDS, imbalanced datasets, sampling algorithms, big data

Procedia PDF Downloads 328

Authors: Anup Anand Deshmukh, Catherine Soladie, Renaud Seguier

Abstract:

Emotion plays a key role in many applications like healthcare, to gather patients’ emotional behavior. Unlike typical ASR (Automated Speech Recognition) problems which focus on 'what was said', it is equally important to understand 'how it was said.' There are certain emotions which are given more importance due to their effectiveness in understanding human feelings. In this paper, we propose an approach that models human stress from audio signals. The research challenge in speech emotion detection is finding the appropriate set of acoustic features corresponding to an emotion. Another difficulty lies in defining the very meaning of emotion and being able to categorize it in a precise manner. Supervised Machine Learning models, including state of the art Deep Learning classification methods, rely on the availability of clean and labelled data. One of the problems in affective computation is the limited amount of annotated data. The existing labelled emotions datasets are highly subjective to the perception of the annotator. We address the first issue of feature selection by exploiting the use of traditional MFCC (Mel-Frequency Cepstral Coefficients) features in Convolutional Neural Network. Our proposed Emo-CNN (Emotion-CNN) architecture treats speech representations in a manner similar to how CNN’s treat images in a vision problem. Our experiments show that Emo-CNN consistently and significantly outperforms the popular existing methods over multiple datasets. It achieves 90.2% categorical accuracy on the Emo-DB dataset. We claim that Emo-CNN is robust to speaker variations and environmental distortions. The proposed approach achieves 85.5% speaker-dependant categorical accuracy for SAVEE (Surrey Audio-Visual Expressed Emotion) dataset, beating the existing CNN based approach by 10.2%. To tackle the second problem of subjectivity in stress labels, we use Lovheim’s cube, which is a 3-dimensional projection of emotions. Monoamine neurotransmitters are a type of chemical messengers in the brain that transmits signals on perceiving emotions. The cube aims at explaining the relationship between these neurotransmitters and the positions of emotions in 3D space. The learnt emotion representations from the Emo-CNN are mapped to the cube using three component PCA (Principal Component Analysis) which is then used to model human stress. This proposed approach not only circumvents the need for labelled stress data but also complies with the psychological theory of emotions given by Lovheim’s cube. We believe that this work is the first step towards creating a connection between Artificial Intelligence and the chemistry of human emotions.

Keywords: deep learning, brain chemistry, emotion perception, Lovheim's cube

Procedia PDF Downloads 154

Authors: Gebreegziabher Hailu Gebrecherkos

Abstract:

This study explores the intersection of categorification and computation within algebraic combinatorial structures, aiming to deepen the understanding of how categorical frameworks can enhance computational methods. We investigate the role of higher-dimensional categories in organizing and analyzing combinatorial data, revealing how these structures can lead to new computational techniques for solving complex problems in algebraic combinatory. By examining examples such as species, posets, and operads, we illustrate the transformative potential of categorification in generating new algorithms and optimizing existing ones. Our findings suggest that integrating categorical insights with computational approaches not only enriches the theoretical landscape but also provides practical tools for tackling intricate combinatorial challenges, ultimately paving the way for future research in both fields.

Keywords: categorification, computation, algebraic structures, combinatorics

Procedia PDF Downloads 14

Authors: Ahmed F. Elaksher, Islam Omar

Abstract:

In this project, we used MOLA-based DEMs to orthorectify HiRISE optical images. The MOLA data was interpolated using the kriging interpolation technique. Corresponding tie points were then digitized from both datasets. These points were employed in co-registering both datasets using GIS analysis tools. Different transformation models, including the affine and projective transformation models, were used with different sets and distributions of tie points. Additionally, we evaluated the use of the MOLA elevations in co-registering the MOLA and HiRISE datasets. The planimetric RMSEs achieved for each model are reported. Results suggested the use of 3D-2D transformation models.

Keywords: photogrammetry, Mars, MOLA, HiRISE

Procedia PDF Downloads 77

Authors: Nabil Shawkat, Jamil Saquer

Abstract:

Social media has accelerated our ability to engage with others and eliminated many communication barriers. On the other hand, the widespread use of social media resulted in an increase in online hate speech. This has drastic impacts on vulnerable individuals and societies. Therefore, it is critical to detect hate speech to prevent innocent users and vulnerable communities from becoming victims of hate speech. We investigate the performance of different deep learning and machine learning algorithms on three different datasets. Our results show that the BERT model gives the best performance among all the models by achieving an F1-score of 90.6% on one of the datasets and F1-scores of 89.7% and 88.2% on the other two datasets.

Keywords: hate speech, machine learning, deep learning, abusive words, social media, text classification

Procedia PDF Downloads 136

Authors: Frederik Calsius, Mirela Popa, Alexia Briassouli

Abstract:

In the field of machine learning, it is common practice to use benchmark datasets to prove the working of a method. The domain of action recognition in videos often uses datasets like Kinet-ics, Something-Something, UCF-101 and HMDB-51 to report results. Considering the properties of the datasets, there are no datasets that focus solely on very short clips (2 to 3 seconds), and on highly-similar fine-grained actions within one specific domain. This paper researches how current state-of-the-art action recognition methods perform on a dataset that consists of highly similar, fine-grained actions. To do so, a dataset of skateboarding tricks was created. The performed analysis highlights both benefits and limitations of state-of-the-art methods, while proposing future research directions in the activity recognition domain. The conducted research shows that the best results are obtained by fusing RGB data with OpenPose data for the Temporal Shift Module.

Keywords: activity recognition, fused deep representations, fine-grained dataset, temporal modeling

Procedia PDF Downloads 231

Authors: Alexander Cho, Harsho Sanyal, Joseph Cataldo

Abstract:

A quantitative analysis of the different variables on both the South and North green roofs of the Jacob K. Javits Convention Center was taken to find mathematical relationships between net radiation and evapotranspiration (ET), average outside temperature, and the lysimeter weight. Groups of datasets were analyzed, and the relationships were plotted on linear and semi-log graphs to find consistent relationships. Antecedent conditions for each rainstorm were also recorded and plotted against the volumetric water difference within the lysimeter. The first relation was the inverse parabolic relationship between the lysimeter weight and the net radiation and ET. The peaks and valleys of the lysimeter weight corresponded to valleys and peaks in the net radiation and ET respectively, with the 8/22/15 and 1/22/16 datasets showing this trend. The U-shaped and inverse U-shaped plots of the two variables coincided, indicating an inverse relationship between the two variables. Cross variable relationships were examined through graphs with lysimeter weight as the dependent variable on the y-axis. 10 out of 16 of the plots of lysimeter weight vs. outside temperature plots had R² values > 0.9. Antecedent conditions were also recorded for rainstorms, categorized by the amount of precipitation accumulating during the storm. Plotted against the change in the volumetric water weight difference within the lysimeter, a logarithmic regression was found with large R² values. The datasets were compared using the Mann Whitney U-test to see if the datasets were statistically different, using a significance level of 5%; all datasets compared showed a U test statistic value, proving the null hypothesis of the datasets being different from being true.

Keywords: green roof, green infrastructure, Javits Center, evapotranspiration, net radiation, lysimeter

Procedia PDF Downloads 114

Authors: Hailah Alballaa, Azeddine Chikh

Abstract:

Several challenges impede the adoption of Recommender Systems for Technology Enhanced Learning (TEL): to collect and identify possible datasets; to select between different recommender approaches; to evaluate their performances. The aim is of this paper is twofold: First, it aims to introduce a survey on the most significant work in this area. Second, it aims at identifying possible research directions.

Keywords: datasets, content-based filtering, recommender systems, TEL

Procedia PDF Downloads 244

Authors: Jacob Green, Cecilia Cabrera, Maximilian Jakobs, Andrea Dimitracopoulos, Mark van der Wilk, Ryan Greenhalgh

Abstract:

Predicting drug properties is key in drug discovery to enable de-risking of assets before expensive clinical trials and to find highly active compounds faster. Interest from the machine learning community has led to the release of a variety of benchmark datasets and proposed methods. However, it remains unclear for practitioners which method or approach is most suitable, as different papers benchmark on different datasets and methods, leading to varying conclusions that are not easily compared. Our large-scale empirical study links together numerous earlier works on different datasets and methods, thus offering a comprehensive overview of the existing property classes, datasets, and their interactions with different methods. We emphasise the importance of uncertainty quantification and the time and, therefore, cost of applying these methods in the drug development decision-making cycle. To the best of the author's knowledge, it has been observed that the optimal approach varies depending on the dataset and that engineered features with classical machine learning methods often outperform deep learning. Specifically, QSAR datasets are typically best analysed with classical methods such as Gaussian Processes, while ADMET datasets are sometimes better described by Trees or deep learning methods such as Graph Neural Networks or language models. Our work highlights that practitioners do not yet have a straightforward, black-box procedure to rely on and sets a precedent for creating practitioner-relevant benchmarks. Deep learning approaches must be proven on these benchmarks to become the practical method of choice in drug property prediction.

Keywords: activity (QSAR), ADMET, classical methods, drug property prediction, empirical study, machine learning

Procedia PDF Downloads 81

Authors: Chih-Fong Tsai, Ya-Han Hu

Abstract:

The aim of feature selection (or dimensionality reduction) is to filter out unrepresentative features (or variables) making the classifier perform better than the one without feature selection. Since there are many well-known feature selection algorithms, and different classifiers based on different selection results may perform differently, very few studies consider examining the effect of performing different feature selection algorithms on the classification performances by different classifiers over different types of datasets. In this paper, two widely used algorithms, which are the genetic algorithm (GA) and information gain (IG), are used to perform feature selection. On the other hand, three well-known classifiers are constructed, which are the CART decision tree (DT), multi-layer perceptron (MLP) neural network, and support vector machine (SVM). Based on 14 different types of datasets, the experimental results show that in most cases IG is a better feature selection algorithm than GA. In addition, the combinations of IG with DT and IG with SVM perform best and second best for small and large scale datasets.

Keywords: data mining, feature selection, pattern classification, dimensionality reduction

Procedia PDF Downloads 669

Authors: Fabio Fabris, Alex A. Freitas

Abstract:

Hierarchical classification is a special type of classification task where the class labels are organised into a hierarchy, with more generic class labels being ancestors of more specific ones. Meta-learning for classification-algorithm recommendation consists of recommending to the user a classification algorithm, from a pool of candidate algorithms, for a dataset, based on the past performance of the candidate algorithms in other datasets. Meta-learning is normally used in conventional, non-hierarchical classification. By contrast, this paper proposes a meta-learning approach for more challenging task of hierarchical classification, and evaluates it in a large number of bioinformatics datasets. Hierarchical classification is especially relevant for bioinformatics problems, as protein and gene functions tend to be organised into a hierarchy of class labels. This work proposes meta-learning approach for recommending the best hierarchical classification algorithm to a hierarchical classification dataset. This work’s contributions are: 1) proposing an algorithm for splitting hierarchical datasets into new datasets to increase the number of meta-instances, 2) proposing meta-features for hierarchical classification, and 3) interpreting decision-tree meta-models for hierarchical classification algorithm recommendation.

Keywords: algorithm recommendation, meta-learning, bioinformatics, hierarchical classification

Procedia PDF Downloads 314

Authors: Fares Hedayatit, Arnauld Joly, Panagiotis Papadimitriou

Abstract:

Many machine learning tasks such as text annotation usually require training over very big datasets, e.g., millions of web documents, that can be represented in a sparse input space. State-of the-art tree-based ensemble algorithms cannot scale to such datasets, since they include operations whose running time is a function of the input space size rather than a function of the non-zero input elements. In this paper, we propose an efficient splitting algorithm to leverage input sparsity within decision tree methods. Our algorithm improves training time over sparse datasets by more than two orders of magnitude and it has been incorporated in the current version of scikit-learn.org, the most popular open source Python machine learning library.

Keywords: big data, sparsely representable data, tree-based models, scalable learning

Procedia PDF Downloads 263

Authors: Mikhail Zarechnev, Bora I. Kumova

Abstract:

A syllogism is a deductive inference scheme used to derive a conclusion from a set of premises. In a categorical syllogisms, there are only two premises and every premise and conclusion is given in form of a quantified relationship between two objects. The different order of objects in premises give classification known as figures. We have shown that the ordered combinations of 3 generalized quantifiers with certain figure provide in total of 108 syllogistic moods which can be considered as different inference rules. The classical syllogistic system allows to model human thought and reasoning with syllogistic structures always attracted the attention of cognitive scientists. Since automated reasoning is considered as part of learning subsystem of AI agents, syllogistic system can be applied for this approach. Another application of syllogistic system is related to inference mechanisms on the Semantic Web applications. In this paper we proposed the mathematical model and algorithm for syllogistic reasoning. Also the model of iterative syllogistic reasoning in case of continuous flows of incoming data based on case–based reasoning and possible applications of proposed system were discussed.

Keywords: categorical syllogism, case-based reasoning, cognitive architecture, inference on the semantic web, syllogistic reasoning

Procedia PDF Downloads 411

Authors: Shahadut Hossain

Abstract:

Measurement errors in continuous covariates and/or misclassifications in categorical covariates are common in epidemiological studies. Regression analysis ignoring such mismeasurements seriously biases the estimated main and interaction effects of covariates on the outcome of interest. Thus, adjustments for such mismeasurements are necessary. In this research, we propose a Bayesian parametric framework for eliminating deleterious impacts of covariate mismeasurements in logistic regression. The proposed adjustment method is unified and thus can be applied to any generalized linear and non-linear regression models. Furthermore, adjustment for covariate mismeasurements requires validation data usually in the form of either gold standard measurements or replicates of the mismeasured covariates on a subset of the study population. Initial investigation shows that adequacy of such adjustment depends on the sizes of main and validation samples, especially when prevalences of the categorical covariates are low. Thus, we investigate the impact of main and validation sample sizes on the adjusted estimates, and provide a general guideline about these sample sizes based on simulation studies.

Keywords: measurement errors, misclassification, mismeasurement, validation sample, Bayesian adjustment

Procedia PDF Downloads 408

Authors: Suzan Arslanturk, Mohammad-Reza Siadat, Theophilus Ogunyemi, Ananias Diokno

Abstract:

Missing data is a common challenge in statistical analyses of most clinical survey datasets. A variety of methods have been developed to enable analysis of survey data to deal with missing values. Imputation is the most commonly used among the above methods. However, in order to minimize the bias introduced due to imputation, one must choose the right imputation technique and apply it to the correct type of missing data. In this paper, we have identified different types of missing values: missing data due to skip pattern (SPMD), undetermined missing data (UMD), and genuine missing data (GMD) and applied rough set imputation on only the GMD portion of the missing data. We have used rough set imputation to evaluate the effect of such imputation on prediction by generating several simulation datasets based on an existing epidemiological dataset (MESA). To measure how well each dataset lends itself to the prediction model (logistic regression), we have used p-values from the Wald test. To evaluate the accuracy of the prediction, we have considered the width of 95% confidence interval for the probability of incontinence. Both imputed and non-imputed simulation datasets were fit to the prediction model, and they both turned out to be significant (p-value < 0.05). However, the Wald score shows a better fit for the imputed compared to non-imputed datasets (28.7 vs. 23.4). The average confidence interval width was decreased by 10.4% when the imputed dataset was used, meaning higher precision. The results show that using the rough set method for missing data imputation on GMD data improve the predictive capability of the logistic regression. Further studies are required to generalize this conclusion to other clinical survey datasets.

Keywords: rough set, imputation, clinical survey data simulation, genuine missing data, predictive index

Procedia PDF Downloads 168

Authors: Armin Nabaei, M. Omair Ahmad, M. N. S. Swamy

Abstract:

Many facial expression and emotion recognition methods in the traditional approaches of using LDA, PCA, and EBGM have been proposed. In recent years deep learning models have provided a unique platform addressing by automatically extracting the features for the detection of facial expression and emotions. However, deep networks require large training datasets to extract automatic features effectively. In this work, we propose an efficient emotion detection algorithm using face images when only small datasets are available for training. We design a deep network whose feature extraction capability is enhanced by utilizing several parallel modules between the input and output of the network, each focusing on the extraction of different types of coarse features with fined grained details to break the symmetry of produced information. In fact, we leverage long range dependencies, which is one of the main drawback of CNNs. We develop this work by introducing a Dynamic Soft-Margin SoftMax.The conventional SoftMax suffers from reaching to gold labels very soon, which take the model to over-fitting. Because it’s not able to determine adequately discriminant feature vectors for some variant class labels. We reduced the risk of over-fitting by using a dynamic shape of input tensor instead of static in SoftMax layer with specifying a desired Soft- Margin. In fact, it acts as a controller to how hard the model should work to push dissimilar embedding vectors apart. For the proposed Categorical Loss, by the objective of compacting the same class labels and separating different class labels in the normalized log domain.We select penalty for those predictions with high divergence from ground-truth labels.So, we shorten correct feature vectors and enlarge false prediction tensors, it means we assign more weights for those classes with conjunction to each other (namely, “hard labels to learn”). By doing this work, we constrain the model to generate more discriminate feature vectors for variant class labels. Finally, for the proposed optimizer, our focus is on solving weak convergence of Adam optimizer for a non-convex problem. Our noteworthy optimizer is working by an alternative updating gradient procedure with an exponential weighted moving average function for faster convergence and exploiting a weight decay method to help drastically reducing the learning rate near optima to reach the dominant local minimum. We demonstrate the superiority of our proposed work by surpassing the first rank of three widely used Facial Expression Recognition datasets with 93.30% on FER-2013, and 16% improvement compare to the first rank after 10 years, reaching to 90.73% on RAF-DB, and 100% k-fold average accuracy for CK+ dataset, and shown to provide a top performance to that provided by other networks, which require much larger training datasets.

Keywords: computer vision, facial expression recognition, machine learning, algorithms, depp learning, neural networks

Procedia PDF Downloads 74

Authors: Akshay Paranjape, Nils Plettenberg, Robert Schmitt

Abstract:

Traditional optimization methods like evolutionary algorithms are widely used in production processes to find an optimal or near-optimal solution of control parameters based on the simulated environment space of a process. These algorithms are computationally intensive and therefore do not provide the opportunity for real-time optimization. This paper utilizes the Deep Reinforcement Learning (DRL) framework to find an optimal or near-optimal solution for control parameters. A model based on maximum a posteriori policy optimization (Hybrid-MPO) that can handle both numerical and categorical parameters is used as a benchmark for comparison. A comparative study shows that DRL can find optimal solutions of similar quality as compared to evolutionary algorithms while requiring significantly less time making them preferable for real-time optimization. The results are confirmed in a large-scale validation study on datasets from production and other fields. A trained XGBoost model is used as a surrogate for process simulation. Finally, multiple ways to improve the model are discussed.

Keywords: reinforcement learning, evolutionary algorithms, production process optimization, real-time optimization, hybrid-MPO

Procedia PDF Downloads 112

Authors: Aakansha Gupta, Neha Vadnere, Tapasvi Soni, M. Anbarsi

Abstract:

Protein secondary structure prediction is one of the most important goals pursued by bioinformatics and theoretical chemistry; it is highly important in medicine and biotechnology. Some aspects of protein functions and genome analysis can be predicted by secondary structure prediction. This is used to help annotate sequences, classify proteins, identify domains, and recognize functional motifs. In this paper, we represent protein secondary structure as a mathematical model. To extract and predict the protein secondary structure from the primary structure, we require a set of parameters. Any constants appearing in the model are specified by these parameters, which also provide a mechanism for efficient and accurate use of data. To estimate these model parameters there are many algorithms out of which the most popular one is the EM algorithm or called the Expectation Maximization Algorithm. These model parameters are estimated with the use of protein datasets like RS126 by using the Bayesian Probabilistic method (data set being categorical). This paper can then be extended into comparing the efficiency of EM algorithm to the other algorithms for estimating the model parameters, which will in turn lead to an efficient component for the Protein Secondary Structure Prediction. Further this paper provides a scope to use these parameters for predicting secondary structure of proteins using machine learning techniques like neural networks and fuzzy logic. The ultimate objective will be to obtain greater accuracy better than the previously achieved.

Keywords: model parameters, expectation maximization algorithm, protein secondary structure prediction, bioinformatics

Procedia PDF Downloads 480

Authors: Nádia Pereira, Paula Ferreira, Sofia Francisco, Sofia Oliveira, Sidclay Souza, Paula Paulino, Ana Margarida Veiga Simão

Abstract:

Social media has progressed into a platform for hate speech among its users, and thus, there is an increasing need to develop automatic detection classifiers of offense and conflicts to help decrease the prevalence of such incidents. Online communication can be used to intentionally harm someone, which is why such classifiers could be essential in social networks. A possible application of these classifiers is the automatic detection of cyberbullying. Even though identifying the aggressive language used in online interactions could be important to build cyberbullying datasets, there are other criteria that must be considered. Being able to capture the language, which is indicative of the intent to harm others in a specific context of online interaction is fundamental. Offense and hate speech may be the foundation of online conflicts, which have become commonly used in social media and are an emergent research focus in machine learning and natural language processing. This study presents two Portuguese language offense-related datasets which serve as examples for future research and extend the study of the topic. The first is similar to other offense detection related datasets and is entitled Aggressiveness dataset. The second is a novelty because of the use of the history of the interaction between users and is entitled the Conflicts/Attacks dataset. Both datasets were developed in different phases. Firstly, we performed a content analysis of verbal aggression witnessed by adolescents in situations of cyberbullying. Secondly, we computed frequency analyses from the previous phase to gather lexical and linguistic cues used to identify potentially aggressive conflicts and attacks which were posted on Twitter. Thirdly, thorough annotation of real tweets was performed byindependent postgraduate educational psychologists with experience in cyberbullying research. Lastly, we benchmarked these datasets with other machine learning classifiers.

Keywords: aggression, classifiers, cyberbullying, datasets, hate speech, machine learning

Procedia PDF Downloads 228