We present new approaches for building yearly and seasonal models for 5-minute ahead electricity load forecasting. They are evaluated using two full years of Australian electricity load data. We first analyze the cyclic nature of the electricity load and show that the autocorrelation function captures these patterns and can be used to extract useful features, as the data is highly linearly correlated. Using the selected feature sets, we then evaluate the predictive performance of four algorithms, representing different prediction paradigms. We found linear regression to be the most accurate and fastest algorithm, outperforming the industry model based on backpropagation neural networks and all baselines. Our results also show that there is no accuracy gain in building models for each season in comparison to building a single yearly model. industry forecasters. The commonly cited advantages of BPNN in comparison to the traditional model-based methods are their ability: 1) to learn from examples without prior assumptions about the process that generated the data, as required in the model-based methods; 2) to form complex non-linear input/output mapping as opposed to the linear mapping formed by the traditional model-based systems. In this paper we examine the need for nonlinear prediction models by taking a closer look at the cyclic nature of the 5- minute electricity load and also by comparing linear and nonlinear prediction algorithms that learn from examples. We also investigate the benefits of building seasonal prediction models as opposed to building a single yearly model. To evaluate the results we use a large dataset of Australian electricity load data. More specifically, our