INVITED SPEAKERS

Although artificial neural network models produce very successful results in time series forecasting, an outlier or outliers in the data set adversely affects the forecasting performance. The dendritic neuron model neural network, which is the most similar neural network model to the artificial neural network model, is also negatively affected by outliers in the data set like many artificial neural network models in the literature. In this study, a robust learning algorithm based on Talwar's M estimator, a data preprocessing method to reduce the effect on the inputs of the network, and a median statistic to prevent the effect of outliers in the output of the network are used together to prevent the dendritic neuron model artificial neural network from being affected by outliers in the data set. In addition, the proposed neural network model is trained with a symbiotic organism search algorithm. In order to evaluate the performance of the proposed method, the time series of the Spanish, Italian and German stock exchanges are analyzed based on their closing prices in certain years. According to the results of the analysis of the time series of the related stock markets both in their original form and by injecting outliers into the time series, the proposed method has superior forecasting performance even when the time series contains outliers and when it does not contain outliers.

Projection pursuit is a multivariate statistical technique aimed at finding interesting low-dimensional data projections. It looks for the data projection which maximizes the projection pursuit index, that is a measure of its interestingness. After an interesting projection is found, it is removed to facilitate the search for other interesting features. Projection pursuit deals with three major challenges of multivariate analysis: the curse of dimensionality, the presence of irrelevant features and the limitations of visual perception. Its applications have been hampered by several difficulties of computational, interpretative and inferential nature. This talk outlines the main features of projection pursuit and illustrates them with the well-known Iris dataset.

Spatial data is data with quantitative and qualitative characteristics that can be associated with specific geographical locations. The research, acquisition, representation, modelling and inference stages of spatial data, which are defined as field data, geostatistical data and point data containing spatial information, constitute spatial data science. Spatial data science can also be defined as the process of spatial data-information discovery. This process includes exploratory data analysis, data visualisation, machine learning algorithms and spatial data analysis. Spatial data analysis is the data analysis that determines the possible relationships between other spatial events with statistical methods that explain the interaction between the attributes that define the data and the processes of the data, taking into account the structure of the data existing in a space. When it comes to statistical analysis of spatial data, spatial data analysis is also called spatial statistics. Spatial statistics is used to support decision-making in strategic decision processes in various fields such as environment, public health, ecology, agriculture, urban planning, economy, society, etc. In this respect, this study aims to provide information on how spatial statistics, which has a wide range of applications today, can be applied, how patterns in spatial data can be revealed in the spatial data-information discovery process, and how spatial statistics can be obtained by using computer programmes (Python, R and/or ArcGIS). In addition, it is aimed to contribute to the studies of researchers on the spatial data science process from a statistical perspective.

Clustering data is a fundamental task in applied data analysis. Under a model-based clustering point of view, finite mixture models serve as an all-purpose workhorse for attacking this problem. Despite their flexibility, the estimation of such models is not trivial and computational methods (frequentist or Bayesian) should be incorporated. However, certain computational and inferential difficulties arise, especially in the case of high dimensional datasets. This talk will review recent advances on these issues and apply the proposed methods in real datasets arising in various scientific disciplines such as immunology, genomics, zoology, demographics and social network data.

He completed his undergraduate education at Akdeniz University, Faculty of Agriculture, Department of Horticulture in 1990. In the same year, he started his master's degree at Yüzüncü Yıl University, Department of Science and Horticulture. He completed his master's degree in the same institute in 1993 and his doctorate in 1997.

In 2001, Yüzüncü Yıl University Faculty of Agriculture, Department of Horticulture was appointed as Assoc. He was appointed as Assoc. In 2004, he received the title of associate professor and in 2005 he was appointed as an associate professor at Tokat Gaziosmanpaşa University, Faculty of Agriculture, Department of Horticulture. In 2010, he was appointed as a professor in the same department and is still working as a professor in the same department.