Churn rate describes the rate at which customers abandon a product or service. Identifying churn-risk customers is essential for telecom sectors to retain old customers and maintain a higher competitive advantage. The purpose of this paper is to explore an effective method for detecting the risk of customer churn in telecom sectors through comparing the advanced machine learning methods and their optimization algorithms. Based on two different telecom datasets, Mutual Information classifier was firstly utilized to select the most critical features relevant to customer churn. Next, the controlled-ratio undersampling strategy was employed to balance both minority and majority classes. Key hyperparameter optimization algorithms of Grid Search, Random Search, and Genetic Algorithms were then combined to fit the three promising machine learning models-Random Forest, Support Vector Machines, and K-nearest neighbors into the customer churn prediction problem. Six evaluation metrics-Accuracy, Recall, Precision, AUC, F1-score and Mean Absolute Error, were last used to evaluate the performance of the proposed models. The experimental results have revealed that the RF algorithm optimized by Grid Search based on a low-ratio undersampling strategy (RF-GS-LR) outperformed other models in extracting hidden information and understanding future churning behaviors of customers on both datasets, with the maximum accuracy of 99 and 95 on the applied dataset 1-2 respectively.

Typhoon forecasting has always been a vital function of the meteorological department. Accurate typhoon forecasts can provide a priori information for the relevant meteorological departments and help make more scientific decisions to reduce the losses caused by typhoons. However, current mainstream typhoon forecast methods are very challenging and expensive due to the complexity of typhoon motion and the scarcity of ocean observation stations. In this paper, we propose a typhoon track prediction model, DeepTyphoon, which integrates convolutional neural networks and long short-term memory (LSTM). To establish the relationship between the satellite image and the typhoon center, we mark the typhoon center on the satellite image. Then, we use hybrid dilated convolution to extract the cloud features of the typhoon from satellite images and use LSTM to predict these features. Finally, we detect the location of the typhoon according to the predictive markers in the output image. Experiments are conducted using 13, 400 satellite images of time series of the Northwest Pacific from 1980 to 2020 and 8420 satellite images of time series of the Southwest Pacific released by the Japan Meteorological Agency. From the experimentation, the mean average error of the 6-hour typhoon prediction result is 64.17 km, which shows that the DeepTyphoon prediction model significantly outperforms existing deep learning approaches. It achieves successful typhoon track prediction based on satellite images.

There are some major difficulties when predicting shelf life of chilled Penaeus vannamei in practice, such as numerous prediction indexes, complicated index measurement methods, and low prediction accuracy. In order to overcome these complications, the Grey Relational Analysis (GRA) was first employed to determine the correlation degree between the quality indexes of chilled Penaeus vannamei and the remaining shelf life. The results have revealed that sensory score and pH value are the two most highly relevant quality indicators of the remaining shelf life of chilled Penaeus vannamei, with the relational degree of 0.8567 and 0.8285, respectively. The sensory score, pH value, and the critical environmental index – temperature were selected as the input parameters of Support Vector Regression (SVR) prediction model. The experimental results showed that the prediction accuracy of SVR prediction model based on GRA is 95.66%, which is similar to that of the SVR model based on all indexes (95.5%) and higher than that of the traditional dynamic model (90.08%). By reducing the dimensions of the prediction indexes, simplifying the measurement methods of indexes, and ensuring relatively high prediction accuracy, the SVR shelf-life prediction model for chilled Penaeus vannamei based on GRA is more feasible for practical use.