Julian Pucheta

A STATISTICALLY DEPENDENT APPROACH FOR THE MONTHLY RAINFALL FORECASTFROM ONE POINT OBSERVATIONS

In this work an adaptive linear filter model in a autoregressive moving average (ARMA) topology for forecasting time series is presented. The time series are composed by observations of the accumulative rainfall every month during several years. The learning rule used to adjust the filter coefficients is mainly based on the gradient-descendent method. In function of the long and short term stochastic dependence of the time series, we propose an on-line heuristic law to set the training process and to modify the filter topology. The input patterns for the predictor filter are the values of the time series after applying a time-delay operator. Hence, the filter’s output will tend to approximate the current value available from the data series. The approach is tested over a time series obtained from measures of the monthly accumulative rainfall from La Perla, Cordoba, Argentina. The performance of the presented approach is shown by forecasting the following 18 months from a hypothetical actual time for four time series of 102 data length.

Time Series Forecasting Using Bayesian Method: Application to Cumulative Rainfall

In this work an algorithm to adjust parameters using a Bayesian method for cumulative rainfall time series forecasting implemented by an ANN-filter is presented. The criterion of adjustment comprises to generate a posterior probability distribution of time series values from forecasted time series, where the structure is changed by considering a Bayesian inference. These are approximated by the ANN based predictor in which a new input is taken in order for changing the structure and parameters of the filter. The proposed technique is based on the prior distribution assumptions. Predictions are obtained by weighting up all possible models and parameter values according to their posterior distribution. Furthermore, if the time series is smooth or rough, the fitting algorithm can be changed to suit, in function of the long or short term stochastic dependence of the time series, an on-line heuristic law to set the training process, modify the NN topology, change the number of patterns and iterations in addition to the Bayesian inference in accordance with Hurst parameter H taking into account that the series forecasted has the same H as the real time series. The performance of the approach is tested over a time series obtained from samples of the Mackey-Glass delay differential equations and cumulative rainfall time series from some geographical points of Cordoba, Argentina.

Rainfall Forecasting Using Sub sampling Nonparametric Methods

This article presents a comparison of two sub sampling nonparametric methods for designing algorithms to forecast time series from the cumulative monthly rainfall. Both approaches are based on artificial feed-forward neural networks ANNs. The main contribution is to divide the rainfall time series forecasting problem using non-parametric methods by subdivision into stages of smoothing, so in this manner the time series are smoothed in order to simplify the prediction problem. The first case depicts an algorithm to forecast high roughness time series that set the parameters of a nonlinear autoregressive model NAR based on ANNs, which uses as a reference the Hurst parameter associated to the time series. The second case, the methodology consists of generating smoothing time series by sampling the time series data, and each individual time series is associated with a predictor filter. Thus, depending on the data, others time series are obtained by sampling with an increasing interval. For each one of the time series generated, a specific ANN-based filter is adjusted, and each one generates a forecast that is then averaged among other subsamples time series, resulting so in a mix of predictor filters. The results are evaluated on high roughness time series from the Mackey Glass Equation MG and from cumulative monthly historical rainfall data from one geographic location. The results are encouraging; deserve study and investment in implementation effort for the geographical locations of interest.

Energy Associated Tuning Method for Short-Term Series Forecasting by Complete and Incomplete Datasets

Abstract This article presents short-term predictions using neural networks tuned by energy associated to series based-predictor filter for complete and incomplete datasets. A benchmark of high roughness time series from Mackay Glass (MG), Logistic (LOG), Henon (HEN) and some univariate series chosen from NN3 Forecasting Competition are used. An average smoothing technique is assumed to complete the data missing in the dataset. The Hurst parameter estimated through wavelets is used to estimate the roughness of the real and forecasted series. The validation and horizon of the time series is presented by the 15 values ahead. The performance of the proposed filter shows that even a short dataset is incomplete, besides a linear smoothing technique employed; the prediction is almost fair by means of SMAPE index. Although the major result shows that the predictor system based on energy associated to series has an optimal performance from several chaotic time series, in particular, this method among other provides a good estimation when the short-term series are taken from one point observations.

Short-term rainfall time series prediction with incomplete data

In order to predict short-term times series with incomplete data, a proposed approach is presented based on the energy associated of series. A benchmark of rainfall time series and Mackay Glass (MG) samples are used. An average smoothing technique is adopted to complete the dataset. The structure of the predictor filter is changed taking into account the energy associated of the short series. The H parameter is used to estimate the roughness of the complete series, the real and forecasted one. The next 15 values are used as validation and horizon of the time series presented by series of cumulative monthly historical rainfall from La Sevillana, Cordoba, Argentina and samples of the Mackay Glass (MG) differential equation. The performance of the proposed filter shows that even the short dataset is incomplete, besides a linear smoothing technique employed, the prediction is almost fair. Although the major result shows that the predictor system based on energy associated to series has an optimal performance from several samples of MG equations and, in particular, MG1.6 and SEV rainfall time series, this method provides a good estimation when the short-term series are taken from one point observations.

Neuro-dynamic programming-based optimal control for crop growth in precision agriculture

The agricultural sector is one activity of the major importance in the Argentinean economy, and their production management and control systems are an important subject of research and development. A neuro-dynamic programming based optimal controller for crop-greenhouse systems is proposed. The neurocontroller drives the crop-growth development minimizing a predefined performance index, which considers minimization of the greenhouse operative costs and the final state errors under physical constraints on process variables and actuator signals. In particular, it is applied to guide the tomato seedling crop development through control of a greenhouse microclimate. In the neurocontroller design process nonlinear dynamic behavior of the crop greenhouse system and the July climate data of 1999 of San Juan, Argentina, are considered. The obtained control law is suboptimal due to the use of neural networks to approximate both the optimal cost-to-go function and optimal policy. In order to show the practical feasibility and performance of the proposed neurocontroller, simulation studies were carried out for the tomato-seedling crop development, which would ease the transition to experimentations on a scale model of a greenhouse available in the Instituto de Automatical Laboratory.

Approximate Optimal Control-Based Neurocontroller with a State Observation System for Seedlings Growth in Greenhouse

In this paper, an approximate optimal control-based neurocontroller for guiding the seedlings growth in greenhouse is presented. The main goal of this approach is to obtain a close-loop operation with a state neurocontroller, whose design is based on approximate optimal control theory. The neurocontroller drives the progress of the crop growth development while minimizing a predefined cost function in terms of operative costs and final state errors under physical constraints on process variables and actuator signals. The aim is to find an approximate optimal control policy to guide the development of tomato seedlings from an initial to a desired state by controlling the greenhouses microclimate. In this paper we propose an indirect measuring of the seedlings growth state using artificial vision. In order to show the performance and practical feasibility of the proposed approach, an experiment was carried out for the development of tomato seedings

A New Approach for Time Series Forecasting: Bayesian Enhanced by Fractional Brownian Motion with Application to Rainfall Series

A new predictor algorithm based on Bayesian enhanced approach (BEA) for long-term chaotic time series using artificial neural networks (ANN) is presented. The technique based on stochastic models uses Bayesian inference by means of Fractional Brownian Motion as model data and Beta model as prior information. However, the need of experimental data for specifying and estimating causal models has not changed. Indeed, Bayes method provides another way to incorporate prior knowledge in forecasting models; the simplest representations of prior knowledge in forecasting models are hard to beat in many forecasting situations, either because prior knowledge is insufficient to improve on models or because prior knowledge leads to the conclusion that the situation is stable. This work contributes with long-term time series prediction, to give forecast horizons up to 18 steps ahead. Thus, the forecasted values and validation data are presented by solutions of benchmark chaotic series such as Mackey-Glass, Lorenz, Henon, Logistic, Rossler, Ikeda, Quadratic one-dimensional map series and monthly cumulative rainfall collected from Despenaderos, Cordoba, Argentina. The computational results are evaluated against several non-linear ANN predictors proposed before on high roughness series that shows a better performance of Bayesian Enhanced approach in long-term forecasting.

A Neuro-Dynamic Programming-Based Optimal Controller for Tomato Seedling Growth in Greenhouse Systems

This work proposes a neuro-dynamic programming-based optimal controller to guide the growth of tomato seedling crops by manipulating its environmental conditions in a greenhouse. The neurocontroller manages the growth development of the crop, while minimizing a predefined cost function that considers the operative costs and the final state errors under physical constraints on process variables and actuator signals. The aim is to guide the growth of tomato seedlings by controlling the microclimate of the greenhouse. The design process of the neurocontroller considers the nonlinear dynamic behavior of the crop-greenhouse system model and the real climate data. Simulations of the proposed approach allow for contrasting its performance against those of other strategies for tomato seedling crop development subject to various climatic conditions.

A New Approach to Image Segmentation with Two-Dimensional Hidden Markov Models

Image segmentation is one of the fundamental problems in computer vision. In this work, we present a new segmentation algorithm that is based on the theory of two-dimensional hidden Markov models (2D-HMM). Unlike most 2D-HMM approaches we do not apply the Viterbi Algorithm, instead we present a computationally efficient algorithm that propagates the state probabilities through the image. This approach can easily be extended to higher dimensions. We compare the proposed method with a 2D-HMM standard algorithm and Iterated Conditional Modes using real world images like a radiography or a satellite image as well as synthetic images. The experimental results show that our approach is highly capable of condensing image segments. This gives our algorithm a significant advantage over the standard algorithm when dealing with noisy images with few classes.