This study investigates the use of several trading strategies, based on Machine Learning methods, to profit on the risk premium of the Nordic electricity base-load week futures. The information set is only composed by financial data from January 02, 2006 to November 15, 2017. The results point out that the Support Vector Machine is the best method, but, most importantly, they highlight that all individual models are valuable, in the sense that their combination provides a robust trading procedure, generating an average profit of at least 26% per year, after considering trading costs and liquidity constraints. The results are robust to the different data partitions, and there is no evidence that the profitability of the trading strategies has decreased in recent years. We claim that this market allows for profitable speculation, namely by using combinations of non-linear signal extraction techniques.

中文翻译：

---

利用机器学习从北欧电力期货的风险溢价中获利

本研究调查了基于机器学习方法的几种交易策略的使用，以从北欧电力基本负荷周期货的风险溢价中获利。该信息集仅由 2006 年 1 月 2 日至 2017 年 11 月 15 日的财务数据组成。结果指出支持向量机是最好的方法，但最重要的是，它们强调所有单个模型都是有价值的，在感觉他们的组合提供了一个强大的交易程序，在考虑交易成本和流动性限制后，平均每年产生至少 26% 的利润。结果对不同的数据分区是稳健的，并且没有证据表明近年来交易策略的盈利能力有所下降。我们声称这个市场允许进行有利可图的投机，