Refinement Model Based on Deep Learning Technique for Prediction of Temperature Using Missing Data

Abstract

Researchers have recently turned their focus to time sequence predicting of meteorological such as daily heat in an effort to overcome the limitations of standard forecasting methods. Due to the difficulty of the task, it is difficult to create and choice an precise time-series forecast perfect. This is a critical factor in human life and many other areas, including agriculture and manufacturing. People's health will be adversely affected by an increase in temperatures in the highland urban heat, especially in the summer, as a result of this. As a result of this paper's research, a novel temperature prediction model based on deep learning has been developed (i.e., the progressive deep cascade categorization model). In order to achieve this, a large volume of high-quality model training data is required. A drawback to weather data collection is the inability to measure data that has been overlooked. There is a high probability of missing or incorrect data due to the nature of data collection. To make up for the lost weather data, the proposed temperature prediction ideal is being used to fine-tune the existing data. Research also uses a deep learning network for time-series data modelling because the temperature changes throughout the year. Various deep learning techniques are also being examined to verify the model's efficacy. In particular, the suggested model's refinement function can be used to restore lost data. The model is retrained using the refined data after all the missing data is refined. Finally, the proposed model for predicting temperature has the capability of doing so. The suggested model's (RMSE) root-mean-squared error, accuracy, precision, and recall are used to evaluate its performance.