Siwei Gao

Theory and method of genetic-neural optimizing cut-off grade and grade of crude ore

Cut-off grade for ore drawing is a kind of technological method used to control the process of drawing in sublevel caving with no sill pillar. The cut-off grade for ore drawing means the grade of ore in the last time (current time) of ore drawing. Grade of crude ore is the grade of ore entering the milling workshop after ore mixing. Cut-off grade and grade of crude ore are key parameters of production and management in mine system. Genetic algorithm and neural networks nesting method are used in this research to simulate the highly complexity and highly non-linear relationship between variables in mining system, to optimize the cut-off grade and grade of crude ore. The idea is detailed as follows. Cut-off grade and grade of crude ore are joined as chromosome of population for evolution computation; Self-adaptive neural network is used to obtain the local connection between the revenue (fitness function) and chromosome; Genetic algorithm is performed to search the optimal cut-off grade and grade of crude ore globally. The inner layer of nesting is neural networks, which is used to compute loss rate, amount of tailing ore and total cost; the outer layer is evolutionary computation, which is used to get the revenue. The inner layer carries out local approximation, and the outer carries out global search. These two layers carry out the optimization of cut-off grade and grade of crude ore jointly. Take Daye Iron Mine as an example, and the result shows that, the present scheme (cut-off grade is 18%, grade of crude ore is 41-43%) should be improved. During the period of August to November in the year 2007, the optimal cut-off grade is 15.8%, and optimal grade of crude ore is 43.7762-44.1387%, the optimized scheme can improve the present value by 9.01-9.44 million yuan.

An Improved Genetic k-means Algorithm for Optimal Clustering

In the classical k-means algorithm, the value of k must be confirmed in advance. It is difficult to confirm accurately the value of k in reality. This paper proposes an improved genetic k-means algorithm (IGKM) and constructs a fitness function defined as a product of three factors, maximization of which ensures the formation of a small number of compact clusters with large separation between at least two clusters. At last, two artificial and three real-life data sets are considered for experiments that compare IGKM with k-means algorithm, GA-based method and genetic k-means algorithm (GKM) by inter-cluster distance (ITD), inner-cluster distance (IND) and rate of separation exactness. The experiments show that IGKM can automatically reach the optimal value of k with high accuracy

A hybrid MPSO-BP structure adaptive algorithm for RBFNs

This paper introduces a novel hybrid algorithm to determine the parameters of radial basis function neural networks (number of neurons, centers, width and weights) automatically. The hybrid algorithm combines the mix encoding particle swarm optimization algorithm with the back propagation (BP) algorithm to form a hybrid learning algorithm (MPSO-BP) for training Radial Basis Function Networks (RBFNs), which adapts to the network structure and updates its weights by choosing a special fitness function. The proposed method is used to deal with three nonlinear problems, and the results obtained are compared with existent bibliography, showing an improvement over the published methods.

Risk finance for catastrophe losses with Pareto-calibrated Lévy-stable severities.

For catastrophe losses, the conventional risk finance paradigm of enterprise risk management identifies transfer, as opposed to pooling or avoidance, as the preferred solution. However, this analysis does not necessarily account for differences between light- and heavy-tailed characteristics of loss portfolios. Of particular concern are the decreasing benefits of diversification (through pooling) as the tails of severity distributions become heavier. In the present article, we study a loss portfolio characterized by nonstochastic frequency and a class of Lévy-stable severity distributions calibrated to match the parameters of the Pareto II distribution. We then propose a conservative risk finance paradigm that can be used to prepare the firm for worst-case scenarios with regard to both (1) the firms intrinsic sensitivity to risk and (2) the heaviness of the severitys tail.

The optimal cluster number of FCM in complex economic systems

When modeling the complex economic systems, a target system often need to be categorized into a certain clusters by FCM. The optimal cluster number often is dependent of the selected cluster validity function, but there are so many validity functions proposed, it is difficult to get the optimal cluster number in real target system. A method to get the optimal cluster number of FCM in real systems is proposed: Presets some reasonable cluster numbers, and then chooses a cluster number as the optimal cluster number by some representative validity functions. Testing on the X30 and Bensaid data sets demonstrates the effectiveness and reliability of the proposed method, and finally gives an experiment on Chinapsilas 31 regions according to the level of science and technology (S&T) progress.

A nonlinear goal-programming-based DE and ANN approach to grade optimization in iron mining

Abstract This study proposes a combined ‘nonlinear goal-programming’-based ‘differential evolution’ (DE) and ‘artificial neural networks’ (ANN) methodology for grade optimization in iron mining production processes. The nonlinear goal-programming model has decision variables of ‘cutoff grade,’ ‘dressing grade’ and ‘concentrate grade,’ with the goals being ‘concentrate output,’ ‘resource utilization rate’ and ‘economic benefit (profit).’ The model, which contains three unknown functions, the ‘loss rate,’ the ‘ore-dressing metal recovery rate’ and the ‘total cost,’ is subsequently converted into an unconstrained optimization problem, to be solved by our integrated DE–ANN approach. DE is used to search for the optimum combination of the cutoff, dressing and concentrate grades, with the crossover rate in the DE analysis being dynamically adjusted within the evolutionary process. The loss rate is calculated by a regression model, whilst the ore-dressing metal recovery rate and the total cost functions are, respectively, calculated using ‘back-propagation’ and ‘radial basis function’ neural networks. We subsequently go on to analyze a case study of the Daye iron mine in China to demonstrate the reliability and efficiency of our proposed approach. Our study provides a novel approach for decision makers to guide production and management in iron mining.

Incentive Contracting with an Independent Underwriter: Does It Benefit Insurers?

Abstract In this article, we propose a model consisting of an insurance distribution channel compensation scheme, paying special attention to the insurer’s choice of distribution system in both single-period and multi-period settings. We find that the risk factor is the key element in both the insurer’s choice of distribution channel and the distribution channel compensation scheme. The advantage of having an independent underwriter is mainly manifested in lines of business that are more risky. Our analysis suggests that a profit-sharing contingent-commission scheme serves as a risk-sharing mechanism and is especially effective with risky business lines.

A hybrid GA-SA-BPNNs for human capital prediction of China regions

Human capital formation and accelerating economic growth is a representative complex system which is not suitable to measure and forecast by classic linear statistical approaches. This paper presents an approach of fusing genetic algorithm (GA), simulated annealing (SA) and error back propagation neural networks (BPNNs) to predict human capital of China regions. Adopting multi-encoding, the GA-SA-BPNNs can simultaneously optimize the hidden nodes, transfer function, weights and bias of BP networks dynamically and adaptively. Furthermore,the most important factors of human capital formation can be identified by selecting input nodes.