Global Mean Temperature (GMT) is a very important variable to be predicted based on its history. The fluctuations in GMT might not be describable by exact mathematical modelling since the underlying Physics is not yet understood fully, and a data-driven approach to this problem is an alternative worth exploring. A variety of data-driven algorithms in Machine Learning (ML) and Deep Learning Neural Networks (DNN) are available for predicting time series. However, advanced algorithms, like DNN are complex to implement and computationally expensive. In this article, only simple ML methods have been evaluated to predict GMT treating it both as a univariate time series and also casting it to a regression problem. The effectiveness of a large set of simpler ML methods along with different data preparation techniques to forecast the GMT and mean value of GMT over a span of years was examined. It was found that some simple methods did as well or better than the more well-known ones showing merit in trying a large bouquet of algorithms as a first step. Forecasts were satisfactory with an RMSE value of around 0.056 on average, with the lowest value of 0.02. RMSE for mean GMT values ranged from 0.00002 to 0.00036. This establishes a benchmark for the more advanced ML models to reach. Some steps of data preparation were shown to be effective. Application of DNN is recommended to examine if that is capable of predicting GMT with greater accuracy.