M. V. Manoj Kumar

Smartphone based emotion recognition and classification

This paper proposes a method that classifies the emotion status of a human being based on ones interactions with the smart phone. Due to one or the other practical limitations, the existing set of emotion recognition methods are difficult to use on daily basis (most of the known methods cause inconvenience to user since they require devices like wearable sensors, camera, or answering a questionnaire). The essence of this paper is to analyze the textual content of the message and user typing behavior to build a classifier that efficiently classifies the future instances. Each observation in the data set consists of 14 features. A machine learning technique called Naive Bayes classifier is applied to construct the classifier. Method proposed is capable of classifying emotions in one of the seven classes (anger, disgust, happy, sad, neutral, surprised, and fear). Experimental result has shown 72% accuracy in classification.

Inter-point envelope based distance moments for offline signature verification

In this paper, inter-point envelope based distance moments are proposed for offline signature verification. Two types of features such as the DC-line and envelope-to-envelope features are investigated from various known reference points. The detailed and high dimensional inter-point distances are used to estimate centralized moments such as the variance, skewness, kurtosis along with mean. The generated moment features are subsequently applied for training the SVM classifier. A newly suggested approach to generate the moment features by separately analyzing the upper and lower envelopes has shown high verification accuracy of 92.73% on the noisy CEDAR signature database.

Prediction of Gallstone Disease Progression Using Modified Cascade Neural Network

Prediction of disease severity is highly essential for understanding the progression of disease and initiating an early diagnosis, which is priceless in treatment planning. A Modified Cascade Neural Network (ModCNN) is proposed for stratification of the patients who may need Endoscopic Retrograde Cholangiopancreatography (ERCP). In this study, gallstone disease (GSD) whose prevalence is increasing in India is considered. A retrospective analysis of 100 patients was conducted and their case history was recorded along with the routine investigations. Using ModCNN, the associated risk factors were extracted for the prediction of disease progression toward severe complication. The proposed model outperformed showing better accuracy with an area under receiver operating characteristic curve (area under ROC curve) of 0.9793, 0.9643, 0.9869, and 0.9768 for choledocholithiasis, pancreatitis, cholecystitis, and cholangitis, respectively, when compared with Artificial Neural Network (ANN) showing an accuracy of 0.884. Hence, the proposed technique can be used to conduct a nonlinear statistical analysis for the better prediction of disease progression and assist in better treatment planning, avoiding future complications.

Simplifying Spaghetti Processes to Find the Frequent Execution Paths

Control-flow discovery algorithms of Process Mining are capable of generating excellent process models until the process is structured (less number of activities and paths connecting between them). Otherwise, process model with Spaghetti structure will be generated. These models are unstructured, incomprehensible and cannot be used for operational support. This paper proposes the techniques for (1) converting Spaghetti (unstructured) process to Lasagna (structured) process, and (2) Identifying the frequent execution paths in the process under consideration.

Recommending an alternative path of execution using an online decision support system

Prediction of disease severity is highly essential for understanding the progression of disease and initiating an alternative path of execution, which is priceless in treatment planning. An online decision support system (ODeSS) is proposed here for stratification of the patients who may need Endoscopic Retrograde Cholangio-Pancreatography (ERCP) and recommend an alternate path of execution. By this an immediate intervention can be avoided. In this study gallstone disease (GSD) whose prevalence is increasing in India is considered. ODeSS is a versatile non-linear information model which clustered the traces based on the duration of its completion. This is a Retrospective analyses of 575 traces. ODeSS applied the technique of longest common subsequence for identifying the sequence of an online execution and discovering to which cluster of variants it may belong. This discovery assist in taking appropriate clinical decision by recommending an alternative path of execution for such cases which may need emergency interventions. ODeSS performance was evaluated using area under receiver operating characteristic curve (area under ROC curve). This showed an accuracy of 0.9653 in prediction. The proposed model was validated using ROC curve in k-fold cross validation. Hence the proposed ODeSS can be used to conduct a non-linear statistical analysis since, the relationships between the predictive variables are not linear. It can be used as a clinical practice to recommend the path of execution. This would assist in better treatment planning, avoiding future complications.

Distilling Lasagna from Spaghetti Processes

If the operational process is flexible, control flow discovery methods in process mining tend to produce Spaghetti (unstructured) models. Spaghetti models generally consist of large number of activities and paths. These models are unstructured, incomprehensible, difficult to analyse, impossible to use during operational support and enhancement. Due The structural complexity of Spaghetti processes majority of techniques in process mining can not be applied on them. There is a at most necessity to design and develop methods for simplifying the structure of Spaghetti process to make them easily understandable and reusable. The methods proposed in this paper concentrates on offering the tools and techniques for analysing the Spaghetti process. The problems addressed in this paper are 1) converting the unstructured Spaghetti to structured and simplified Lasagna process, 2) identifying the list of possible, significant, and impossible paths of execution in Lasagna process. The proposed technique is verified and validated on real-life road traffic fine management event-log taken from standard repository.

On predicting the frequent execution patterns in information systems

Process mining research discipline offers a spectrum of techniques for analysing event logs. Event logs represent the history of process execution. This information can be used for monitoring, analysing and improving the operational processes. The currently available methods in process mining emphasise on constructing the static process model. These models depict various dimensions of the process under analysis. But, models can only represent the past execution history and cant be used to guide and control the prospectus execution of the process. There is a need for the methods and techniques which guide the future execution of process in the light of recorded information. This paper introduces a technique for identifying and predicting the frequent control-flow execution patterns in information systems. The proposed Position Weight Matrix proven to be efficient during experimentation and validation studies.

Chord oriented gap feature for offline signature verification

In this paper, we address offline signature verification by proposing a new Partial Invariant Chord Oriented Gap (PICOG) feature. The new heuristically developed feature is conceptualized after observing the directional variation of the gaps (sequence of white pixels) between signature strokes. A set of unique and partial invariant chords is identified using genuine and forgery training signatures in the writer dependent system. Different sets of PICOG chords are selected for each writer by defining a threshold (djnv). The similarity threshold (dth) is computed by performing another training step using the PICOG chords. A majority score based approach is selected to determine if the testing signature is genuine or forgery. A maximum accuracy of 82.27% is obtained on the widely used and publicly available, noisy signature database (CEDAR).