1,720,996 research outputs found
Detecting and interpreting the consensus ranking based on the weighted Kemeny distance
No full textThis paper outlines a way for finding the consensus ranking minimizing the sum of the weighted Kemeny distance, using positional weights. The weighted Kemeny distance, introduced by Garc ́ıa-Lapresta and Perez-Rom ́ an, meets the original ́ Kemeny-Snell axioms and it is fully applicable in treating weak orderings. A differential evolution algorithm is ad-hoc defined in order to detect the consensus ranking, namely that ranking that best represents the preferences expressed by a set of individuals
Dynamic modelling of price expectations
No full textEvaluation surveys are often repeated over time in order to check for
trends in subjects’ behaviors and opinions. The paper proposes a dynamic model for
the serial correlation of ratings’ intrinsic components, which is discussed on the basis
of time series of price expectations in Italy collected within a survey organized by
ISTAT
INFLUENCE OF OUTLIERS ON CLUSTER CORRESPONDENCE ANALYSIS
No full textThis paper focuses on determining the influence of outliers on a joint dimension reduction and clustering method for categorical data, namely Cluster Cor- respondence Analysis (CCA). Joint methods, such as CCA, solutions consist of both a cluster membership vector and a set of low dimensional scores for observations and attributes. We evaluate the impact of outliers on the identification of the cluster struc- ture. As a benchmark, we use the tandem approach, which is a sequential application of multiple correspondence analysis followed by K-means clustering. The appraisal is based on synthetic data and outliers generated using an evolutionary algorithm that provides data with a user-defined cluster structure
Quantile composite-based path modeling to estimate the conditional quantiles of health indicators
No full textQuantile Composed-based Path Modeling complements the classical PLS Path Modeling. The latter is widely used to model relationships among latent variables and between the manifest variables and their corresponding latent variables. Since it essentially exploits classical least square regressions, PLS Path Modeling focuses on the effect the predictors exert on the conditional means of the different outcome variables involved in models. Quantile Composed-based Path Modeling extends the analysis to the whole conditional distributions of the outcomes. This paper proposes a procedure to estimate the conditional quantiles for the manifest variables of the outcome blocks. Starting from the information related to a grid of conditional quantiles, it is possible to define the most accurate model for each health indicator and the best predictive model for each Italian province. The proposed method is shown in action both on artificial and real data. The real data concerns the prediction of health indicators
Evaluation of the web usability of the University of Cagliari portal: an eye tracking study
No full textA web portal is one of the main tools used by companies, institutions and individual citizens to make information available to anyone. Designing a portal that has good usability means allowing an average user to find the information he needs as soon as possible. The objective of this work is to evaluate the web usability of the portal of the University of Cagliari, using the eye tracking technology. High school and university students were asked to perform specific tasks within the portal. The results were evaluated through a quantitative analysis of the time and number of fixations required to complete each task, as well as a qualitative analysis of heat maps and gaze plots representing participants' fixations. The analysis has allowed to (i) detect a high efficiency for most of the web pages, (ii) highlight the most critical elements of the portal and (iii) suggest the most appropriate changes to be made
STABILITY OF JOINT DIMENSION REDUCTION AND CLUSTERING
Get PDFSeveral methods for joint dimension reduction and cluster analysis of categorical, continuous or mixed-type data have been proposed over time. These methods combine dimension reduction (PCA/MCA/PCAmix) with partitioning clus- tering (K-means) by optimizing a single objective function. Cluster stability assess- ment is a critical and inadequately discussed topic in the context of joint dimension reduction and clustering. We introduce a resampling scheme that combines boot- strapping and a measure of cluster agreement to assess global cluster stability of joint dimension reduction and clustering solutions and a Jaccard similarity approach for empirical evaluation of the stability of individual clusters. Both approaches are imple- mented in the R package clustrd
MACHINE LEARNING MODELS FOR FORECASTING STOCK TRENDS
No full textThis research addresses the problem of predicting the trends of two
stocks and two stock indexes for the American stock market. In this study, the predictive performance of four machine learning models, are compared. The models investigated include Artificial Neural Networks (ANN), Support Vector Machine (SVM),
Random Forest and Naive-Bayes. Supervised models training is performed through a
10-fold CV approach repeated 3 times, using 10 of the main indicators and oscillators
of technical analysis as input. The experiments conducted show that among the 4,
the Naive-Bayes model gives the worst predictive performance, the Random Forest
obtains discrete results, while the SVM and the ANN are the best performing models
- …
