34 research outputs found
NATIONAL FOREST ASSESSMENT PROGRAM LEBANON DATA ANALYSIS REPORT BY GENANE YOUNESS BEYDOUN (FAO DATA ANALYSIS EXPERT): NATIONAL FOREST AND TREE RESOURCES ASSESSMENT 2003-05 (TCP/LEB/2903)
No full textTCP/LEB/2903 National Forest and Tree Inventory and Assessment of Lebanon became active in July 2003. The main objective of the project was to reinforce the capacity of the Directorate of Rural Development and Natural Resources (DRDNR) in collecting, compiling, analyzing and disseminating reliable and up–to-date information on the forest and trees outside the forest (TOF) resources of Lebanon through training of the national staff on forest and tree inventory (MOA/FAO 2003). It provides the first experiences with systematic field sampling of the forest and TOF resources of Lebanon. This report presents the results obtained in the statistical data analysis of the National Forest Assessment of Lebanon during TCP/LEB/2903
Sur des indices de comparaison de deux classifications
No full textInternational audienceOn étudie la ressemblance entre deux partitions sur les mêmes individus à l'aide du coefficient de corrélation vectorielle RV et du kappa de Cohen (dans le cas où les partitions ont le même nombre de classes). On montre que le RV s'identifie à l'indice J de Janson et Vegelius. On étudie la distribution d'échantillonnage de ces indices pour des paires de partitions proches (issues d'un modèle de classes latentes) afin de donner des valeurs critiques sous des hypothèses réalistes
An empirical comparison between PLS, LASSO, Elasticnet and other models for highly correlated data
No full textInternational audienceWe compare seven regression techniques to solve the question of muticollinearity in multiple regression: PCR, PLS regression, with Ridge, Lasso, Lars, Adaptive Lasso and Elasticnet. An application on real data providing from the Lebanese national center for scientific research CNRSL(Centre National de la Recherche Scientifique Liban) on the area of fire in Lebanon forests in 2005 will be made. The comparison of these techniques is discussed along with some of their advantages and disadvantages
Use of PLS path modeling to study customer satisfaction of two Lebanese mobile operators
No full textInternational audienceThe purpose of this study is to measure customer satisfaction of two mobile operators in Lebanon, Alfa and MTC Touch, from a sample of the telephone customers for these two operators; We draw 500 phone active numbers randomly selected from their system database. After telephone contacts, we were able to record and completed 600 questionnaires referred to the EPSI (European Performance Satisfaction Index) questionnaire made on 23 questions and distributed equally between the two operators; the data collection was carried out between May 2012 and June 2012. The goal is to construct the structural equations modeling and to estimate the measure of customer satisfaction using the ECSI [1] model by PLS approach and build a satisfaction index for each operator. In our work, we discuss the ECSI model (European Customer Satisfaction Index). Indeed, this model solves the main difficulties in measuring customer satisfaction. It provides best reliable results and do better understand and monitor the quality of customer relationship business. The hypothesis are tested and discussed
An Explainable Artificial Intelligence Approach for Remaining Useful Life Prediction
No full textInternational audiencePrognosis and health management depend on sufficient prior knowledge of the degradation process of critical components to predict the remaining useful life. This task is composed of two phases: learning and prediction. The first phase uses the available information to learn the system’s behavior. The second phase predicts future behavior based on the available information of the system and estimates its remaining lifetime. Deep learning approaches achieve good prognostic performance but usually suffer from a high computational load and a lack of interpretability. Complex feature extraction models do not solve this problem, as they lose information in the learning phase and thus have a poor prognosis for the remaining lifetime. A new prepossessing approach is used with feature clustering to address this issue. It allows for restructuring the data into homogeneous groups strongly related to each other using a simple architecture of the LSTM model. It is advantageous in terms of learning time and the possibility of using limited computational capabilities. Then, we focus on the interpretability of deep learning prognosis using Explainable AI to achieve interpretable RUL prediction. The proposed approach offers model improvement and enhanced interpretability, enabling a better understanding of feature contributions. Experimental results on the available NASA C-MAPSS dataset show the performance of the proposed model compared to other common methods
Some measures of agreement between close partitions
No full textAbstract. In order to measure the similarity between two partitions coming from the same data set, we study extensions of the RV-coefficient, the kappa coefficient proposed by Cohen (in case of partitions with same number of classes), and the D2 coefficient proposed by Popping. We find that the RV coefficient is identical to the Janson and Vegelius index. We compare the result coming from kappa’s coefficient to the ordination given by correspondence analysis. We study the empirical distribution of these indices under the hypotheses of a common partition. For this purpose, we use data coming from a latent profile model to formulate the null hypothesis. Keywords. RV, Janson and Vegelius index, Cohen’s Kappa, Popping ’ D2, Correspondence analysis, latent clas
Comparing partitions of two sets of units based on the same variables
No full textRand index, Redundancy index, Discriminant analysis, Latent classes, Partitions, 62H30,
Concordance entre deux partitions: quelques propositions et expériences
No full text8èmes rencontres de la Société Francophone de Classification, Pointe à Pitre, Décembre 2001International audienceNous proposons une méthodologie pour étudier la distribution du critère de Rand, dans le cas où les deux partitions devraient concorder.Utilisant un modèle de profils latents, on simule des données que l'on classe selon deux groupes de variables.La distribution de l'indice de Rand et d'un indice adapté du test de Mac Nemar se révèle être bimodale
Contributions à une méthodologie de comparaison de partitions
No full textPARIS-BIUSJ-Thèses (751052125) / SudocPARIS-BIUSJ-Physique recherche (751052113) / SudocSudocFranceF
Une Méthodologie pour la Comparaison de Partitions
No full textInternational audienceWe propose a methodology for comparing two partitions of the same data set. The study begins by presenting some measures of comparison: the Rand's measure of association, the asymmetric, and the corrected one. We also study two other indices: the first is based on an adaptation of Mac Nemar's test, the second being Jaccard's index. We present the logic form based on pair comparisons for this indexes as well as their sampling distribution under the assumption of independence. Stability of classes is studied by simulating data coming from a latent profile model which is a particular modal of a mixture distribution and we partition them according to 2 groups of variables.Nous proposons une méthodologie pour comparer des partitions d'un même ensemble de données. Nous présentons tout d'abord quelques mesures de comparaison de deux classifications d'un même ensemble de données : l'indice de Rand, sous sa forme brute ou corrigée, ainsi que sa version asymétrique, puis deux autres indices : le premier est inspiré du test de Mac Nemar et le second de l'indice de Jaccard. On présente les écritures logiques et relationnelles de ces indices ainsi que leurs distributions d'échantillonnage sous une hypothèse nulle d'ab-sence de liaison. Pour étudier la stabilité des classes on utilise ensuite un modèle particulier de mélanges de distributions, les profils latents qui sert à simuler des données que l'on classe selon deux groupes de variable
