Forecasting air quality index (AQI) is critically important to provide a basis for government policy makers, especially in public health, smart transportation, energy management, economic development, and sustainable environments. In reality, AQI consists of various components, such as PM2.5, PM10, CO, NO2, and SO2. Although numerous methods have been presented, few studies concurrently considered the causalities of socioeconomic indicators and meteorological factors and different data granularities. The aggregate AQI of Taiwan comprises five representative cities: Taipei, Hsinchu, Taichung, Tainan, and Kaohsiung. Research findings identify seasonal factors, carbon power generation, steel and metal production, highway cargo load, the number of registered cars, and retail and manufacturing employment population as the key indicators to predict the monthly AQI of Taiwan. For the daily AQI of Hsinchu and the hourly AQI of Kaohsiung, PM2.5, PM10, O3, ambient temperature, humidity, wind speed, wind direction, and pollutants (CO, NO2, and SO2) are recognized. Deep learning significantly outperforms machine learning in the hourly AQI while it performs slightly better in the daily AQI. With the presented framework, governments can balance the trade‐offs between economic development and environmental sustainability.

Retail firms are the best representatives of a developed country’s economic condition because they sell many of the necessary goods used for daily consumption, including food, clothes, shoes, electric appliances, and office supplies. This study presents a novel framework to help retail practitioners achieve the following goals: (1) predict sales revenues by identifying significant economic indicators, (2) estimate stable equilibriums by capturing interactive dynamics between competing firms, and (3) derive operational efficiencies and indicate required improvements by conducting performance assessments. To verify the validity of the research, data pertaining to Walmart, Costco, and Kroger are collected. Specifically, the least absolute shrinkage and selection operator (Lasso) is adopted in order to identify significant economic indicators. Consumer price index and regular wage are two common indicators that affect the the three firms’ sales numbers. In sales forecasting, support vector regression (SVR) and multivariate adaptive regression splines (MARS), respectively, perform the best in the training set and the testing set. Finally, the Lotka–Volterra model (LVM) and data envelopment analysis (DEA) are used for competitive analysis and performance assessment. A relationship of economic mutualism has been identified between the three firms. Furthermore, research findings show that Kroger performs inefficiently, though it can expect to increase sales more than the others in stable equilibriums.

The prediction of remaining useful life (RUL) is a critical issue in many areas, such as aircrafts, ships, automobile, and facility equipment. Although numerous methods have been presented to address this issue, most of them do not consider the impacts of feature engineering. Typical techniques include the wrapper approach (using metaheuristics), the embedded approach (using machine learning), and the extraction approach (using component analysis). For simplicity, this research considers feature selection and feature extraction. In particular, principal component analysis (PCA) and sliced inverse regression (SIR) are adopted in feature extraction while stepwise regression (SR), multivariate adaptive regression splines (MARS), random forest (RF), and extreme gradient boosting (XGB) are used in feature selection. In feature selection, the original 15 sensors can be reduced to only four sensors that accumulate more than 80% degrees of importance and not seriously decrease the predictive performances. In feature extraction, only the top three principal components can account for more than 80% variances of original 15 sensors. Further, PCA combined with RF is more recommended than PCA and CNN (convolutional neural network) because it can achieve satisfactory performances without incurring tedious computation.