Introduction: Suitable definition and realization of functional form of variables is required for a multivariate data analysis where in the number of parameters must be finite. In the study with Incomplete data such as survival studies, performing an analysis is constraint to some assumptions, which may be not fulfilling in practical situations. In these cases, artificial neural network (ANN) is recommended which is flexible and distribution free.Aim: This study aim to make a comparison between the predictions of ANN and Weibull regression via a simulation study and based on a real data set example.Material and Method: At first, three random variables was generated from binomial and standardized normal distributions for simulation study. In addition, Weibull survival times were generated based on dependence structure of parameters and independent variables.Afterward, the data was randomly divided into two parts: training and testing data sets.Finally, network performance was assessed by using of least square error of prediction and Bayesian information criterion.Results: Concordance index of ANN and Weibull regression was calculated as 0.812 and 0.785, respectively.Conclusion: The accuracy of ANN prediction was better than that of Weibull prediction.

Recommender systems have been widely used in e-commerce applications. They are a sub-class of information filtering system used to either predict whether a user will prefer an item (prediction problem) or identify a set of k items that will be of user-interest (Top-k recommendation problem). Demanding sufficient ratings to make robust predictions and suggesting qualified recommendations are two significant challenges in recommender systems. However, the latter is far from satisfactory because human decisions are affected by environmental conditions, and they might change over time. In this paper, we introduce an innovative method to impute ratings to missed components of the rating matrix. We also design an ensemble-based method to obtain Top-k recommendations. In order to evaluate the performance of the proposed method, several experiments have been conducted based on 10-fold cross-validation over real-world datasets. The experimental results show that the proposed method is superior to the state-of-the-art competing methods regarding the applied evaluation metrics.

Feature selection is a critical step in machine learning, especially when dealing with high-dimensional and Incomplete data. Traditional methods often struggle with missing values, which are common in real-world applications. This paper introduces Neural Network Feature Selection (NNFS), a novel deep learning-based approach that effectively identifies important features even in the presence of missing data. We provide a variety of comparisons to evaluate the suggested algorithm over existing methods. We demonstrate the accuracy, speed and sensitivity to missed data. According to numerical results, the proposed algorithm outperforms existing methods especially for medium size datasets. Both random and real tests are presented to make the results more realistic.

It is well known that damaged structural members may considerably alter the behaviour of the structures. Careful observation of these changes has often been viewed as a means to identify and assess the location and severity of damages in structures. Among the responses of a structure, natural frequencies and mode shapes are both relatively easy to obtain and independent from external excitation, and therefore, can be used as a measure of the structural behaviour before and after an extreme event which might have led to damage in the structure. This paper applies Charged System Search algorithm to the problem of damage detection using vibration data. The objective is to identify the location and extent of multi-damage in a structure. Both natural frequencies and mode shapes are used to form the required objective function. To moderate the effect of noise on measured data, a penalty approach is applied. Varity of numerical examples including beams, frames and trusses are examined, and the results show that the present methodology can reliably identify damage scenarios using noisy measurements and Incomplete data.

The neoclassical growth model is tested by use of panel data procedure in this research. In the econometric test, simoultanously time series and cross detection will be compared on the basis of panel data method through which their observed points increase and consequently the estimation efficiency will be increased. The examination of neoclassical growth theory has been done with reference to external & internal factors of 52 selected countries from 1960 to 2000. The independent variable of model has been selected on the basis of the result of previous research which explains the result in three separate models: developed countries, developing countries, and whole countries. These factors are such as: Gross National Products with lag of period, work force age, growth rate, education level, the change of capital accumulation and economic trade volum. The consequences of this research is that: neoclassical growth model can explain the major part of economic growth of the countries with use of internal variables. Also with the use of panel procedure of fixed effect, we can see the fundamental differences and structure of the growth process for different countries; and show how the economic, and social conditions affect on the growth.

Durbin's rank test is widely used for testing treatment effects in Balanced Incomplete Block Designs (BIBDs) which have wide applications in sensory analysis. This test is failed for BIBDs when ties data occur. An adjusted version of Durbin rank test for this kind of data is given to solve this problem. Chi-square approximation, which is commonly used for this test, is not adequate for small BIBDs. For this case, we investigate permutation approach for adjusted Durbin rank test. Also, in this study the tests used in BIBDs are compared by simulation study for tied data, which have not been discussed in the sensory literature.