A Time-Series Water Level Forecasting Model Based on Imputation and Variable Selection Method

Time series outlier removal and imputing methods based on Colombian weather stations data

The time data series of weather stations are a source of information for floods. The study of the previous wintertime series allows knowing the behavior of the variables and the result that will be applied to analysis and simulation models that feed variables such as flow and level of a study area. One of the most common problems is the acquisition and transmission of data from weather stations due to atypical values and lost data; this generates difficulties in the simulation process. Consequently, it is necessary to propose a numerical strategy to solve this problem. The data source for this study is a real database where these problems are presented with different variables of weather. This study is based on comparing three methods of time series analysis to evaluate a multivariable process offline. For the development of the study, we applied a method based on the discrete Fourier transform (DFT), and we contrasted it with methods such as the average and linear regression without uncertainty parameters to complete missing data. The proposed methodology entails statistical values, outlier detection, and the application of the DFT. The application of DFT allows the time series completion, based on its ability to manage various gap sizes and replace missing values. In sum, DFT led to low error percentages for all the time series (1% average). This percentage reflects what would have likely been the shape or pattern of the time series behavior in the absence of misleading outliers and missing data.

FM-GRU: A Time Series Prediction Method for Water Quality Based on seq2seq Framework

Water quality prediction plays a crucial role in both enterprise management and government environmental management. However, due to the variety in water quality data, inconsistent frequency of data acquisition, inconsistency in data organization, and volatility and sparsity of data, predicting water quality accurately and efficiently has become a key problem. This paper presents a recurrent neural network water quality prediction method based on a sequence-to-sequence (seq2seq) framework. The gate recurrent unit (GRU) model is used as an encoder and decoder, and a factorization machine (FM) is integrated into the model to solve the problem of high sparsity and high dimensional feature interaction in the data, which was not addressed by the water quality prediction models in prior research. Moreover, due to the long period and timespan of water quality data, we add a dual attention mechanism to the seq2seq framework to address memory failures in deep learning. We conducted a series of experiments, and the results show that our proposed method is more accurate than several typical water quality prediction methods.

Data analysis of Covid-19 pandemic and short-term cumulative case forecasting using machine learning time series methods

The Covid-19 pandemic is the most important health disaster that has surrounded the world for the past eight months. There is no clear date yet on when it will end. As of 18 September 2020, more than 31 million people have been infected worldwide. Predicting the Covid-19 trend has become a challenging issue. In this study, data of COVID-19 between 20/01/2020 and 18/09/2020 for USA, Germany and the global was obtained from World Health Organization. Dataset consist of weekly confirmed cases and weekly cumulative confirmed cases for 35 weeks. Then the distribution of the data was examined using the most up-to-date Covid-19 weekly case data and its parameters were obtained according to the statistical distributions. Furthermore, time series prediction model using machine learning was proposed to obtain the curve of disease and forecast the epidemic tendency. Linear regression, multi-layer perceptron, random forest and support vector machines (SVM) machine learning methods were used. The performances of the methods were compared according to the RMSE, APE, MAPE metrics and it was seen that SVM achieved the best trend. According to estimates, the global pandemic will peak at the end of January 2021 and estimated approximately 80 million people will be cumulatively infected.