Improvements in Weather Forecasting Technique Using Cognitive Internet of Things

Kaushlendra Yadav, Anuj Kumar Singh, Arvind Kumar Tiwari

Computer Systems Science and Engineering · 01 January 2023

In one sentence. Stacked ensemble of ML models on IoT sensor data reports 99.25% accuracy for local weather parameters, and 85% on Australian rainfall. Limited generality.

Abstract

Forecasting the weather is a challenging task for human beings because of the unpredictable nature of the climate. However, effective forecasting is vital for the general growth of a country due to the significance of weather forecasting in science and technology. The primary motivation behind this work is to achieve a higher level of forecasting accuracy to avoid any damage. Currently, most weather forecasting work is based on initially observed numerical weather data that cannot fully cover the changing essence of the atmosphere. In this work, sensors are used to collect real-time data for a particular location to capture the varying nature of the atmosphere. Our solution can give the anticipated results with the least amount of human engagement by combining human intelligence and machine learning with the help of the cognitive Internet of Things. The Authors identified weather-related parameters such as temperature, humidity, wind speed, and rainfall and then applied cognitive data collection methods to train and validate their findings. In addition, the Authors have examined the efficacy of various machine learning algorithms by using them on both data sets i.e., pre-recorded metrological data sets and live sensor data sets collected from multiple locations. The Authors noticed that the results were superior on the sensor data. The Authors developed ensemble learning model using stacked method that achieved 99.25% accuracy, 99% recall, 99% precision, and 99% F1-score for Sensor data. It also achieved 85% accuracy, 86% recall, 85% precision, and 86% F1 score for Australian rainfall data.

Labels

Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.