Shelvin Chand

Multi-objective cooperative coevolution of neural networks for time series prediction

The use of neural networks for time series prediction has been an important focus of recent research. Multi-objective optimization techniques have been used for training neural networks for time series prediction. Cooperative coevolution is an evolutionary computation method that decomposes the problem into subcomponents and has shown promising results for training neural networks. This paper presents a multi-objective cooperative coevolutionary method for training neural networks where the training data set is processed to obtain the different objectives for multi-objective evolutionary training of the neural network. We use different time lags as multi-objective criterion. The trained multi-objective neural network can give prediction of the original time series for preprocessed data sets distinguished by their time lags. The proposed method is able to outperform the conventional cooperative coevolutionary methods for training neural networks and also other methods from the literature on benchmark problems.

Cooperative coevolution of feed forward neural networks for financial time series problem

Intelligent financial prediction systems guide investors in making good investments. Investors are continuously on the hunt for better financial prediction systems. Neural networks have shown good results in the area of financial prediction. Cooperative coevolution is an evolutionary computation method that decomposes the problem into subcomponents and has shown promising results for training neural networks. This paper presents a computational intelligence framework for financial prediction where cooperative coevolutionary feedforward neural networks are used for predicting closing market prices for companies listed on the NASDAQ stock exchange. Problem decomposition is an important step in cooperative co-evolution that affects its performance. Synapse and Neuron level are the main problem decomposition methods in cooperative coevolution. These two methods are used for training neural networks on the given financial prediction problem. The results show that Neuron level problem decomposition gives better performance in general. A prototype of a mobile application is also given for investors that can be used on their Android devices.

Fast Heuristics for the Multiple Traveling Thieves Problem

The traveling thief problem (TTP) is fast gaining attention for being a challenging combinatorial optimization problem. A number of algorithms have been proposed for solving this problem in the recent past. Despite being a challenging problem, it is often argued if TTP is realistic enough because of its formulation, which only allows a single thief to travel across hundreds or thousands of cities to collect (steal) items. In addition, the thief is required to visit all cities, regardless of whether an item is stolen there or not. In this paper we discuss the shortcomings of the current formulation and present a relaxed version of the problem which allows multiple thieves to travel across different cities with the aim of maximizing the groups collective profit. A number of fast heuristics for solving the newly proposed multiple traveling thieves problem (MTTP) are also proposed and evaluated.

Evaluation of co-evolutionary neural network architectures for time series prediction with mobile application in finance

Graphical abstractDisplay Omitted HighlightsIn this paper, we evaluate the performance of coevolutionary feedforward and recurrent neural networks architectures for chaotic time series problems.We further apply them for financial prediction problems selected from the NASDAQ stock exchange.We highlight the challenges in real-time implementation and present a mobile application framework for financial time series prediction.The results, in general, show that recurrent neural networks have better generalisation ability when compared to feedforward networks for real-world time series problems. The fusion of soft computing methods such as neural networks and evolutionary algorithms have given a very promising performance for time series prediction problems. In order to fully harness their strengths for wider impact, effective real-world implementation of prediction systems must incorporate the use of innovative technologies such as mobile computing. Recently, co-evolutionary algorithms have shown to be very promising for training neural networks for time series prediction. Cooperative coevolution decomposes a problem into subcomponents that are evolved in isolation and cooperation typically involves fitness evaluation. The challenge has been in developing effective subcomponent decomposition methods for different neural network architectures. In this paper, we evaluate the performance of two problem decomposition methods for training feedforward and recurrent neural networks for chaotic time series problems. We further apply them for financial prediction problems selected from the NASDAQ stock exchange. We highlight the challenges in real-time implementation and present a mobile application framework for financial time series prediction. The results, in general, show that recurrent neural networks have better generalisation ability when compared to feedforward networks for real-world time series problems.

Multi-objectiveness in the single-objective traveling thief problem

Multi-component problems are optimization problems that are composed of multiple interacting sub-problems. The motivation of this work is to investigate whether it can be better to consider multiple objectives when dealing with multiple interdependent components. Therefore, the Travelling Thief Problem (TTP), a relatively new benchmark problem, is investigated as a bi-objective problem. The results indicate that a multi-objective approach can produce solutions to the single-objective TTP variant while being competitive to current state-of-the-art solvers.

Finding robust solutions for resource constrained project scheduling problems involving uncertainties

Resource constrained project scheduling problem (RCPSP) is a well known problem in the area of discrete optimization. It involves scheduling a given set of activities such that they are completed within minimum possible time, while satisfying a given set of precedence and resource constraints. RCPSP has a wide applicability in a number of industries, such as engineering, management, software, etc. While the classical RCPSP has been extensively studied, literature is rather scarce when it comes to finding robust solutions to RCPSP involving uncertainties. A robust solution in this context is one whose performance is not likely to vary significantly in presence of uncertainties which are inevitable in real life scenarios, such as delays in a particular activity and/or change in the available resources. Towards addressing this gap, in this paper we formulate a variant of RCPSP with stochastic activity durations and resource availability. Further, we propose a simple population based algorithm which aims to find solutions (activity lists) with minimum average makespan in the presence of uncertainties. The output from the algorithm is compared against a chosen optimal solution for the original RCPSP in terms of robustness. We study the performance of the proposed algorithm on a number of different J30 instances taken from the widely used Project Scheduling Library (PSPLib) in order to demonstrate the utility of the approach.

Evolving heuristics for the resource constrained project scheduling problem with dynamic resource disruptions

Abstract Dynamic changes and disruptions are encountered frequently in the domain of project scheduling. The nature of these dynamic events often requires project managers to make quick decisions with regards to effectively re-scheduling the activities. Priority heuristics have a significant potential for such applications due to their simplicity, intuitiveness and low computational cost. In this research, we focus on automated evolution of priority heuristics using a genetic programming hyper-heuristic (GPHH). The proposed approach uses a multi-objective scheme (MO-GPHH) to evolve priority heuristics that can perform better than the existing rules, and at the same time have low complexity. Furthermore, unlike the existing works on evolving priority heuristics that focus on only static problems, this study covers both static and dynamic instances. The proposed approach is tested on a practical dynamic variant of the classical resource constrained project scheduling problem (RCPSP) in which the resource availability varies with time and knowledge about these changes and disruptions only become available as the project progresses. Extensive numerical experiments and benchmarking are performed to demonstrate the efficacy of the proposed approach.