Durga Prasad Muni

Genetic programming for simultaneous feature selection and classifier design

This paper presents an online feature selection algorithm using genetic programming (GP). The proposed GP methodology simultaneously selects a good subset of features and constructs a classifier using the selected features. For a c-class problem, it provides a classifier having c trees. In this context, we introduce two new crossover operations to suit the feature selection process. As a byproduct, our algorithm produces a feature ranking scheme. We tested our method on several data sets having dimensions varying from 4 to 7129. We compared the performance of our method with results available in the literature and found that the proposed method produces consistently good results. To demonstrate the robustness of the scheme, we studied its effectiveness on data sets with known (synthetically added) redundant/bad features.

Texture Generation for Fashion Design Using Genetic Programming

We present a methodology to generate textures for fashion design using genetic programming (GP). The proposed GP based scheme evolves tree representation of procedures to generate textures. We use Contrast of the generated textures/images to filter out poor textures. After filtering, the fitness value of a new texture is set as the fitness value of a cluster of (already generated) textures which is more similar to this new texture. For this, we execute a clustering step during the evolution. Statistical features are used to find the similarity between textures. Since the quality of a texture is best assessed by a human being, if the generated texture is quite dissimilar to the existing textures then users discretion is sought to assign a fitness value to it by visual inspection of the texture

Clustering and Labeling IT Maintenance Tickets

The goal of a Service System in an organization is to deliver uninterrupted service towards achieving business success. Ticketing system is an example of a Service System which is responsible for handling huge volumes of tickets generated by large enterprise IT (Information Technology) infrastructure components and ensuring smooth operation. Instead of manual screening one needs to extract information automatically from them to gain insights to improve operational efficiency. To ensure better operation we propose a framework to cluster incident tickets based on their textual context that can eliminate manual classification of them, which is labor intensive and costly. Further we label each of the clusters by generating meaningful keywords as logical itemsets, extracting candidate labels from Wikipedia articles, and finally scoring each of labels against each cluster. These labels can reflect an adequate and concise specification of each cluster. Further we experiment our approach with industrial ticket data from three different domains and report on the learned experience. We believe that our framework for clustering and labeling will enable enterprises to prioritize the issues in their IT infrastructure and improve the reliability and availability of their services.

Evolution of fuzzy classifiers using genetic programming

In this paper, we propose a genetic programming (GP) based approach to evolve fuzzy rule based classifiers. For a c-class problem, a classifier consists of c trees. Each tree, Ti, of the multi-tree classifier represents a set of rules for class i. During the evolutionary process, the inaccurate/inactive rules of the initial set of rules are removed by a cleaning scheme. This allows good rules to sustain and that eventually determines the number of rules. In the beginning, our GP scheme uses a randomly selected subset of features and then evolves the features to be used in each rule. The initial rules are constructed using prototypes, which are generated randomly as well as by the fuzzy k-means (FKM) algorithm. Besides, experiments are conducted in three different ways: Using only randomly generated rules, using a mixture of randomly generated rules and FKM prototype based rules, and with exclusively FKM prototype based rules. The performance of the classifiers is comparable irrespective of the type of initial rules. This emphasizes the novelty of the proposed evolutionary scheme. In this context, we propose a new mutation operation to alter the rule parameters. The GP scheme optimizes the structure of rules as well as the parameters involved. The method is validated on six benchmark data sets and the performance of the proposed scheme is found to be satisfactory.

Fuzzy QoS Modeling of IT Maintenance Tickets

Ticketing system is an example of a Service System (SS) which is responsible for handling huge volumes of tickets generated by large enterprise IT (Information Technology) infrastructure components, and ensuring smooth operation. An issue is captured as summary on the ticket and once a ticket is resolved, the solution is also noted down on the ticket as resolution. Further the system maintains the provision of recording the time when a ticket is opened, acknowledged to user, resolved and/or closed, from which different QoS parameters could be obtained. For example, Resolution Time can be computed as the difference of resolution date and opening date of the ticket. QoS parameters are used to measure the performance of different aspects of a service. In case of impreciseness of observations of these parameters fuzzy sets seems to be an optimal tool to model them. To ensure better operation for services based on these QoS values we propose a two-stage analysis framework for QoS prediction of incoming tickets which includes fuzzy clustering of incident tickets based on QoS values and building a fuzzy regression model using this categorization and the textual contents of tickets. Further we carry out a fuzzy correlation analysis of different categories (clusters) of QoS parameters. Lastly we report on our experimental results.

Recommending resolutions of ITIL services tickets using Deep Neural Network

Application development and maintenance is a good example of Information Technology Infrastructure Library (ITIL) services in which a sizable volume of tickets are raised everyday for different issues to be resolved in order to deliver uninterrupted service. An issue is captured as summary on the ticket and once a ticket is resolved, the solution is also noted down on the ticket as resolution. It will be beneficial to automatically extract information from the description of tickets to improve operations like identifying critical and frequent issues, grouping of tickets based on textual content, suggesting remedial measures for them etc. In particular, the maintenance people can save a lot of effort and time if they have access to past remedial actions for similar kind of tickets raised earlier based on history data. In this work we propose an automated method based on deep neural networks for recommending resolutions for incoming tickets. We use ideas from deep structured semantic models (DSSM) for web search for such resolution recovery. We project a small subset of existing tickets in pairs and an incoming ticket to a low dimensional feature space, following which we compute the similarity of an existing ticket with the new ticket. We select the pair of tickets which has the maximum similarity with the incoming ticket and publish both of its resolutions as the suggested resolutions for the latter ticket. The experiment of our data sets shows that we are able to achieve a promising similarity match of about 70% - 90% between the suggestions and the actual resolution.

Boosting Aided Approaches to QoS Prediction of IT Maintenance Tickets

Ticketing system is an example of a Service System (SS) which is responsible for handling huge volumes of tickets generated by large enterprise IT (Information Technology) infrastructure components, and ensuring smooth operation. The system maintains the provision of recording the time that reflects when a ticket is opened, acknowledged to user, resolved and/or closed, from which different QoS parameters could be obtained. For example, Resolution Time can be computed as the difference of resolution date and opening date of the ticket. One needs to use new technology solutions in QoS-related analysis like categorization of tickets according to their QoS, predicting QoS parameters for new tickets etc., to improve the performance of the SS. In this work we propose boosting oriented solutions to QoS prediction of tickets using crisp and fuzzy set models of QoS. In particular, we employ a two-stage analysis framework for QoS prediction for incoming tickets which includes clustering incident tickets based on QoS values and building a regression model using this categorization and the textual contents of tickets. We carry out experiments on industrial data sets using different techniques for prediction. We improve the quality of prediction by using suitable boosting techniques. We propose random forest boosting on Logistic Regression and gradient boosting on NNLS for our purpose, both of which improve the performance of prediction. We report these results and compare them.

A novel rewards-based protocol and decision-making technique for transshipment in supply chains

We propose a novel rewards based protocol and online decision-making technique to transshipment between stocking locations each supplied periodically to replenish their stocks. Our approach enables independent decision-making by the stocking locations while each of them can adopt replenishment strategies as they wish. The relevant trade-off in our context is that of a significant reduction in the shortage costs for a smaller investment in transport costs. We conduct a multi-period simulation study where in each period transshipment decisions are followed by demand occurrence followed by demand satisfaction.

A computational framework for learning production planning policies

In a variety of production settings in industries such as consumer packaged goods (CPG), pharma and other consumer goods, a number of related items are produced together in the same production mode. The production of these families require setup of capacities, which can incur significant costs in terms of labor, material, etc. Hence production planners are faced with the challenge of determining production quantities and the sequence of changeovers so as to achieve an optimal long-term cost-containment strategy. In this paper, we detail the problem formulation and a simulation based policy learning framework. We also discuss the results of our computations performed using this framework.