Prediction Research And Case Analysis In The Context Of Big Data

In the era of big data,a variety of factors（particularly meteorological factors）have been applied to PM_2.5 concentration prediction,revealing a clear discrepancy in timescale.To capture the complicated multi-scale relationship with PM_2.5-related factors,a novel multi-factor&multi-scale method is proposed for PM_2.5 forecasting.Three major steps are taken:（1）multi-factor analysis,to select predictive factors via statistical tests;（2）multi-scale analysis,to extract scale-aligned components via multivariate empirical mode decomposition;and（3）PM_2.5 prediction,including individual prediction at each timescale and ensemble prediction across different timescales.The first empirical study focuses on the PM_2.5 of Cangzhou,which is one of the most air-polluted cities in China,and indicates that the proposed multifactor&multi-scale learning paradigms statistically outperform their corresponding original techniques（without multi-factor and multi-scale analysis）,semi-improved variants（with either multi-factor or multi-scale analysis）,and similar counterparts（with other multi-scale analyses）in terms of prediction accuracy.The second empirical study focuses on the PM_2.5 in the Indian capital of New Delhi,and details the multi-factor&multi-scale learning paradigms that considers Google Trends.Furthermore,the proposed learning paradigms are compared with other benchmark models,i.e.,their corresponding the original type（without multi-factor and multi-scale analysis）,the semi-improved variants without Google Trends（with either multi-factor or multi-scale analysis）,and the semi-improved variant considering Google Trends（with either multi-factor or multi-scale analysis）and the multi-factor&multi-scale method without Google Trends made a systematic comparative analysis,the empirical results show that:the multi-factor&multi-scale model proposed in this paper,which takes into account the Google Trends,has achieved good results in PM_2.5 prediction.