Predicting Global Average Land Temperatures Using Microsoft Azure Machine Learning Studio
Sagar Tewari1, Sarthak Agarwal2 and Madhulika Bhatia3*
Published: May 30, 2022
To accurately calculate the global average temperature has proven to be an arduous task since the 19th century, the major reasons are maintaining accurate records of the same locations over a lengthy period which has strained the meteorologists, especially in places located remotely, like mountains or deserts. This is the reason meteorologists use global averages which generally span over 3 decades to give perspective and context to information.
We decided to use the Linear Regression Model to achieve this task at hand. Linear regression is an algorithm which is used for determining the relationship between an dependent variable and some independent variables which are also known as scalar response explanatory variables respectively. The relationship between dependent and independent variables are calculated using predictor functions which are linear in nature whose parameters which are not known are found out through the means of the data provided. Hence, it is paramount that we give our model a clean and digestible data so that it learns well and predicts accurately. It’s equally important that we maintain the dimensions of the data and be careful that it is not too high or not too low, it should be just right. We also employed the Pearson’s Correlation method for Feature Selection. After which we also tuned the hyperparameters among other things to gain the most suitable parameters for the model. Finally, we compared our results from normal model generation to hyperparameter tuned model.