scispace - formally typeset
Search or ask a question
Author

M. Shashi

Bio: M. Shashi is an academic researcher from Andhra University. The author has contributed to research in topics: Cluster analysis & Fuzzy clustering. The author has an hindex of 8, co-authored 37 publications receiving 420 citations.

Papers
More filters
Journal ArticleDOI
TL;DR: Non linear regression method is found to be suitable to train the SVM for weather prediction and the results are compared with Multi Layer Perceptron (MLP) trained with back-propagation algorithm and the performance of SVM is finding to be consistently better.
Abstract: —Weather prediction is a challenging task for researchers and has drawn a lot of research interest in the recent years. Literature studies have shown that machine learning techniques achieved better performance than traditional statistical methods. This paper presents an application of Support Vector Machines (SVMs) for weather prediction. Time series data of daily maximum temperature at a location is analyzed to predict the maximum temperature of the next day at that location based on the daily maximum temperatures for a span of previous n days referred to as order of the input. Performance of the system is observed over various spans of 2 to 10 days by using optimal values of the kernel function. Non linear regression method is found to be suitable to train the SVM for this application. The results are compared with Multi Layer Perceptron (MLP) trained with back-propagation algorithm and the performance of SVM is found to be consistently better.

281 citations

Journal ArticleDOI
TL;DR: In this paper, two techniques are proposed to generate session passwords using text and colors which are resistant to shoulder surfing and suitable for Personal Digital Assistants.
Abstract: Textual passwords are the most common method used for authentication. But textual passwords are vulnerable to eves dropping, dictionary attacks, social engineering and shoulder surfing. Graphical passwords are introduced as alternative techniques to textual passwords. Most of the graphical schemes are vulnerable to shoulder surfing. To address this problem, text can be combined with images or colors to generate session passwords for authentication. Session passwords can be used only once and every time a new password is generated. In this paper, two techniques are proposed to generate session passwords using text and colors which are resistant to shoulder surfing. These methods are suitable for Personal Digital Assistants.

66 citations

Journal ArticleDOI
TL;DR: A framework is introduced for identification of news articles related to top trending topics/hashtags and multi-document summarization of unifiable news articles based on the trending topics for capturing opinion diversity on those topics.
Abstract: Vectorization is imperative for processing textual data in natural language processing applications. Vectorization enables the machines to understand the textual contents by converting them into meaningful numerical representations. The proposed work targets at identifying unifiable news articles for performing multi-document summarization. A framework is introduced for identification of news articles related to top trending topics/hashtags and multi-document summarization of unifiable news articles based on the trending topics, for capturing opinion diversity on those topics. Text clustering is applied to the corpus of news articles related to each trending topic to obtain smaller unifiable groups. The effectiveness of various text vectorization methods, namely the bag of word representations with tf-idf scores, word embeddings, and document embeddings are investigated for clustering news articles using the k-means. The paper presents the comparative analysis of different vectorization methods obtained on documents from DUC 2004 benchmark dataset in terms of purity.

55 citations

Journal ArticleDOI
TL;DR: Manifold alignment approach of TL that transforms the source and target domains into a common latent space to evade the problem of different feature spaces and different marginal probability distributions among the domains is applied.

32 citations

Journal Article
TL;DR: A survey of the state of art research on periodic pattern mining algorithms and their application areas was given and a discussion of merits and demerits of these algorithms was given.
Abstract: Owing to a large number of applications periodic pattern mining has been extensively studied for over a decade. Periodic pattern is a pattern that repeats itself with a specific period in a give sequence. Periodic patterns can be mined from datasets like biological sequences, continuous and discrete time series data, spatiotemporal data and social networks. Periodic patterns are classified based on different criteria. Periodic patterns are categorized as frequent periodic patterns and statistically significant patterns based on the frequency of occurrence. Frequent periodic patterns are in turn classified as perfect and imperfect periodic patterns, full and partial periodic patterns, synchronous and asynchronous periodic patterns, dense periodic patterns, approximate periodic patterns. This paper presents a survey of the state of art research on periodic pattern mining algorithms and their application areas. A discussion of merits and demerits of these algorithms was given. The paper also presents a brief overview of algorithms that can be applied for specific types of datasets like spatiotemporal data and social networks.

18 citations


Cited by
More filters
01 Jan 2002

9,314 citations

Journal Article
TL;DR: In this article, the authors explore the effect of dimensionality on the nearest neighbor problem and show that under a broad set of conditions (much broader than independent and identically distributed dimensions), as dimensionality increases, the distance to the nearest data point approaches the distance of the farthest data point.
Abstract: We explore the effect of dimensionality on the nearest neighbor problem. We show that under a broad set of conditions (much broader than independent and identically distributed dimensions), as dimensionality increases, the distance to the nearest data point approaches the distance to the farthest data point. To provide a practical perspective, we present empirical results on both real and synthetic data sets that demonstrate that this effect can occur for as few as 10-15 dimensions. These results should not be interpreted to mean that high-dimensional indexing is never meaningful; we illustrate this point by identifying some high-dimensional workloads for which this effect does not occur. However, our results do emphasize that the methodology used almost universally in the database literature to evaluate high-dimensional indexing techniques is flawed, and should be modified. In particular, most such techniques proposed in the literature are not evaluated versus simple linear scan, and are evaluated over workloads for which nearest neighbor is not meaningful. Often, even the reported experiments, when analyzed carefully, show that linear scan would outperform the techniques being proposed on the workloads studied in high (10-15) dimensionality!.

1,992 citations

Journal ArticleDOI
TL;DR: In this paper, the authors synthesize multiple methods for machine learning (ML) model interpretation and visualization (MIV) focusing on meteorological applications, which has recently exploded in popularity.
Abstract: This paper synthesizes multiple methods for machine learning (ML) model interpretation and visualization (MIV) focusing on meteorological applications. ML has recently exploded in popularit...

268 citations

Proceedings ArticleDOI
10 Aug 2015
TL;DR: This work studies specifically the power of making predictions via a hybrid approach that combines discriminatively trained predictive models with a deep neural network that models the joint statistics of a set of weather-related variables.
Abstract: Weather forecasting is a canonical predictive challenge that has depended primarily on model-based methods. We explore new directions with forecasting weather as a data-intensive challenge that involves inferences across space and time. We study specifically the power of making predictions via a hybrid approach that combines discriminatively trained predictive models with a deep neural network that models the joint statistics of a set of weather-related variables. We show how the base model can be enhanced with spatial interpolation that uses learned long-range spatial dependencies. We also derive an efficient learning and inference procedure that allows for large scale optimization of the model parameters. We evaluate the methods with experiments on real-world meteorological data that highlight the promise of the approach.

219 citations