Stock index series is Non-stationary, Nonlinear and factors with impact on stock index fluctuation are complex, a time series forecasting model combined ARIMA model and wavelet neural network is presented. The combined model uses BP neural network as the main framework, uses wavelet basis function instead of transfer function in the network, also add some inner factors of the time series mining by ARIMA model, as the part impute of Wavelet Neural Network. So it is more scientific and rational that using inner factors and external other factors. The last simulate experiment shows that the wavelet neural network forecasting model based on ARIMA has higher accuracy than ARIMA model or BP network.

The main significance of utilizing the high-performance concrete as an effective item in the construction industry is the compressive strength assessment which requires a vast investigation of the design mix with calculated relevant compressive strength. Through the intelligence approaches, planning an accurate relationship between high-performance concrete different mix designs and their compressive strength is obtainable with the lowest cost of time and finance. In this regard, two models based on support vector regression methods are developed. The optimal output is calculated by tuning support vector regression key constraints by flow direction and biography-based optimization algorithm. The data set collected from the literature is divided into the training, and the testing phase, where the training data is used to develop the models, and the testing data is utilized to validate the accuracy of the models. The results showed a higher accuracy of the FDA_SVR method than the BBO_SVR method, with R2 values of 0.9939 and 0.9755, respectively. moreover, the U_95 confidence level obtained about 4.0273 and 7.367 for the FDA-SVR and BBO-SVR, respectively, demonstrating the accurate prediction capacity of the FDA-SVR model for high-performance concrete compressive strength.

In this paper, an improved feature selection algorithm by conditional mutual information with Parzen window was proposed, which adopted conditional mutual information as an evaluation criterion of feature selection in order to overcome the deficiency of feature redundant and used Parzen window to estimate the probability density functions and calculate the conditional mutual information of continuous variables, in such a way as to achieve feature selection for continuous data.