Chin-Yuan Fan

A neural network with a case based dynamic window for stock trading prediction

Stock forecasting involves complex interactions between market-influencing factors and unknown random processes. In this study, an integrated system, CBDWNN by combining dynamic time windows, case based reasoning (CBR), and neural network for stock trading prediction is developed and it includes three different stages: (1) screening out potential stocks and the important influential factors; (2) using back propagation network (BPN) to predict the buy/sell points (wave peak and wave trough) of stock price and (3) adopting case based dynamic window (CBDW) to further improve the forecasting results from BPN. The system developed in this research is a first attempt in the literature to predict the sell/buy decision points instead of stock price itself. The empirical results show that the CBDW can assist the BPN to reduce the false alarm of buying or selling decisions. Nine different stocks with different trends, i.e., upward, downward and steady, are studied and one individual stock (AUO) will be studied as case example. The rates of return for upward, steady, and downward trend stocks are higher than 93.57%, 37.75%, and 46.62%, respectively. These results are all very promising and better than using CBR or BPN alone.

Data clustering and fuzzy neural network for sales forecasting: A case study in printed circuit board industry

In order to obtain a better control of market trend and profit for the company, timely identification of sales is very important for businesses. Upward and downward trends in sales signify new market trends and understanding of sales trends is important for marketing as well as for customer retention. This research develops a hybrid model by integrating K-mean cluster and fuzzy neural network (KFNN) to forecast the future sales of a printed circuit board factory. Based on the K-mean clustering technique, the historical data can be classified into different clusters. The accuracy of the forecasted model can be further improved by referring the new data to be forecasted from a more focused region, i.e., a smaller region after clustering. Numerical data of various affecting factors and actual demand of the past 5 years of the printed circuit board (PCB) factory are collected and input into the hybrid model for future monthly sales forecasted. The experimental results derived from the proposed model show the effectiveness of the hybrid model when compared with other approaches.

Integrating a Piecewise Linear Representation Method and a Neural Network Model for Stock Trading Points Prediction

Recently, the piecewise linear representation (PLR) method has been applied to the stock market for pattern matching. As such, similar patterns can be retrieved from historical data and future prices of the stock can be predicted according to the patterns retrieved. In this paper, a different approach is taken by applying PLR to decompose historical data into different segments. As a result, temporary turning points (trough or peak) of the historical stock data can be detected and inputted to the backpropagation neural network (BPN) for supervised training of the model. After this, a new set of test data can trigger the model when a buy or sell point is detected by BPN. An intelligent PLR (IPLR) model is further developed by integrating the genetic algorithm with the PLR to iteratively improve the threshold value of the PLR. Thus, it further increases the profitability of the model. The proposed system is tested on three different types of stocks, i.e., uptrend, steady, and downtrend. The experimental results show that the IPLR approach can make significant amounts of profit on stocks with different variations. In conclusion, the proposed system is very effective and encouraging in that it predicts the future trading points of a specific stock.

An ensemble of neural networks for stock trading decision making

Stock turning signals detection are very interesting subject arising in numerous financial and economic planning problems. In this paper, Ensemble Neural Network system with Intelligent Piecewise Linear Representation for stock turning points detection is presented. The Intelligent piecewise linear representation method is able to generate numerous stocks turning signals from the historic data base, then Ensemble Neural Network system will be applied to train the pattern and retrieve similar stock price patterns from historic data for training. These turning signals represent short-term and long-term trading signals for selling or buying stocks from the market which are applied to forecast the future turning points from the set of test data. Experimental results demonstrate that the hybrid system can make a significant and constant amount of profit when compared with other approaches using stock data available in the market.

Integrating a Piecewise Linear Representation Method with Dynamic Time Warping System for Stock Trading Decision Making

Stock turning points detection is a very interesting subject arising in numerous financial and economic planning problems. In this paper, a piecewise linear representation method with dynamics time warping system for stock turning points detection is presented. The piecewise linear representation method is able to generate numerous stocks turning points from the historic data base, then the dynamic time warping system will be applied to retrieve similar stock price patterns from historic data for training the system. These turning points represent short-term trading signals for selling or buying stocks from the market. A Back-Propagation neural network (B.P.N) is further applied to learn the connection weights from these historic turning points and afterwards it is applied to forecast the future turning points from the set of test data. Experimental results demonstrate that the system integrating PLR and neural networks can make a significant amount of profit when compared with other approaches using stock data.

Evolving Neural Network with Dynamic Time Warping and Piecewise Linear Representation System for Stock Trading Decision Making

Stock turning points detection is a very interesting subject arising in numerous financial and economic planning problems. In this paper, Evolving Neural Network model with Dynamic Time warping Piecewise Linear Representation system for stock turning points detection is presented. The piecewise linear representation method is able to generate numerous stocks turning points from the historic data base, then Evolving Neural Network model will be applied to train the pattern and retrieve similar stock price patterns from historic data for training the system. These turning points represent short-term trading signals for selling or buying stocks from the market. it is applied to forecast the future turning points from the set of test data. Experimental results demonstrate that the hybrid system integrating DPLR and evolving neural networks can make a significant and constant amount of profit when compared with other approaches using stock data.

The Hybrid Model Development of Clustering and Back Propagation Network in Printed Circuit Board Sales Forecasting

Reliable prediction of sales can improve the quality of business strategy. This research develops a hybrid model by integrating K-mean cluster and Back Propagation Network (KBPN) to forecast the future sales of a printed circuit board factory. Base on the K-mean clustering technique, the history data can be classified into different clusters, thus the noise of the original data can be reduced and a more accurate prediction model can be established. Numerical data of various affecting factors and actual demand of the past 5 years of the printed circuit board (PCB) factory are collected and input into the hybrid model for future monthly sales forecasting. Experimental results show the effectiveness of the hybrid model when comparing it with other approaches.

Data Clustering and Fuzzy Neural Network for Sales Forecasting in Printed Circuit Board Industry

Reliable prediction of sales can improve the quality of business strategy. This research develops a hybrid model by integrating K-mean cluster and fuzzy back propagation network (KFBPN) to forecast the future sales of a printed circuit board factory. Based on the K-mean clustering technique, the historic data can be classified into different clusters, thus the noise of the original data can be reduced and a more homogeneous region can be established for a more accurate prediction. Numerical data of various affecting factors and actual demand of the past 5 years of the printed circuit board (PCB) factory are collected and input into the hybrid model for future monthly sales forecasting. Experimental results show the effectiveness of the hybrid model when compared with other approaches

A Depth-First Mutation-Based Genetic Algorithm for Flow Shop Scheduling Problems

This paper presents a novel memetic genetic algorithm (GA) for the flow shop scheduling problem by combining mutation-based local search with traditional genetic algorithm. The local search is based on the depth-first mutation-based searching process and the depth, i. e., the number of total mutation within each generation is according to the number of jobs to be scheduled. In traditional GA, the optimal solution may just next to the current best one however the combination of crossover and mutation may generate individuals with the solution jumping off the optimal zones. Therefore, in this research the classical mutation is replaced by depth-first multiple mutations within each generation. The multi-mutation can provide a more completely deep searching during each generation therefore there are more chances for the evolving searching procedure to reach to the optimal zone. In addition, the SA based acceptance rate is designed to be incorporated into the searching procedure; therefore the convergence rate of the hybrid GA can be further improved. The test problems are selected from the OR library, and the computational results show that the hybrid GA has a better solution quality than simple GA and NEH heuristic

Develop a sub-population Memetic Algorithm for multi-objective scheduling problems

Memetic Algorithm is a population-based approach for heuristic search in optimization problems. It has shown that this mechanic performs better than traditional Genetic Algorithms for some problem. In order to apply in the multi-objective problem, the basic local search heuristics are combined with crossover operator in the sub-population in this research. This approach proposed is named as Sub-population with Memetic Algorithm, which is applied to deal with multi-objective Flowshop Scheduling Problems. Besides, the Artificial Chromosome with probability matrix will be introduced when the algorithm evolves to certain iteration for injecting to individual to search better combination of chromosomes, this mechanism will make faster convergent time for evolving. Compares with MOSA, the experiments result show that this algorithm possess fast convergence and average scatter of Pareto solutions simultaneously for solving multi-objective Flowshop Scheduling Problems in test instances.