A smoothing conjugate gradient method for a kind of generalized tensor complementarity problem

Jingjing Sun, Yuanyuan Chen

Abstract: The complementarity problem is widely used in engineering, game and traffic equilibrium. In this paper, a kind of generalized tensor complementarity problem is studied. The generalized tensor complementarity problem is transformed into a system of nonsmooth equations by using complementary function, and then it is equivalently transformed into an unconstrained optimization problem by smoothing the nonsmooth equations. Finally, a smoothing conjugate gradient method is presented. The global convergence of the proposed method is proved under general assumptions, and the effectiveness of the proposed method is also verified by numerical examples.

: Download (free)  CSE-PDF

Analysis and forecast of the development trend of Building Integrated Photovoltaic plate index

Li Wu , Shupei Du , Dong liu , Jieyu Ding

Abstract: Building-Integrated Photovoltaic (BIPV) plays an important role in the realization of "carbon peak" and "carbon neutralization" in China. Therefore, it is important to study the BIPV plate index to predict future trends and make relevant policies. Firstly, some valuable data which can effectively reflect the stock trend are screened and preprocessed. Define a calculation formula of a plate index and a moving average line of the plate index, using entropy weight method, we select the index that can depict the market value of each stock to calculate the corresponding weight of each stock, and then get the plate index and moving average price, and draw the correlative image. Then, divide the historical data into a training set and a test set (8:2), process the features through feature engineering, use the Gradient Boosting Decision Tree model (GBDT) to learn the historical sector index data, and predict the trading day index. The predicted data and the calculated real data are subjected to error analysis and parameter adjustment to obtain a fitted model, and draw the correlation coefficient diagram between the BIPV sector index and the Shanghai Stock Exchange index to obtain the correlation coefficient between them. Finally, take the average value of VaR value of each stock within three months as the risk index of the stock, sort the stocks, and establish a portfolio optimization model: mean-variance portfolio model, give the optimal investment plan, adopt PEST The analysis method analyzes the future development trend of China's BIPV industry.

: Download (free)  CSE-PDF

Multi-index quantitative stock selection strategy based on machine learning

Shupei Du , Li Wu , Jieyu Ding

Abstract: With the gradual development and improvement of China's financial market, investors are more and more keen to predict and analyze stock prices. Compared with traditional fundamental technical analysis, modern analytical trading tools such as quantitative trading and machine learning based on big data show unique advantages. This paper selects CSI A-Shares excluding all ST shares in all trading days in 2021 for research, uses stock volume price technical indicators-opening price, closing price, MACD, MA, turnover, volume ratio, etc., to screen the daily stock pool, and then uses machine learning algorithms such as XGBoost to predict the rise and fall of the selected stocks, and finally introduces activity indicators such as rising speed, trading volume, capital inflow and so on. Combined with the Topsis comprehensive evaluation method to score the stock, select the three stocks with the highest score to simulate buying, and calculate the rate of return within five trading days after the position. At the end of the paper the model is summarized and prospected.

: Download (free)  CSE-PDF

The property of expected residual minimization method for stochastic tensor complementarity problem

Mengdie Lu

Abstract: This paper focus on a deterministic formulation for the stochastic tensor complementarity problem, which called the expected residual minimization formulation. The expected residual minimization formulation aims at minimizing an expected residual defined by nonlinear complementarity functions. The expected residual minimization function based on the min function is proved to be a semi-smooth function. The property of the expected residual minimization function based on the min function at strictly feasible points is also proved. The related examples are given to illustrate the property of the ERM function.

: Download (free)  CSE-PDF