无线传感器网络的节点运行往往受限于能量供给。对太阳能进行采集并转换成电能存储,可以延长节点的使用寿命。对太阳能进行能量预测,可以更好地规划和使用采集到的能量,这有助于节省能源、避免浪费,提升无线传感器网络的生存周期。针对太阳能预测,提出一种基于自回归积分移动平均-长短期记忆(Autoregressive Integrated Moving Average-Long Short Term Memory,ARIMA-lstm)组合模型的能量预测方法。首先,采用ARIMA模型来对太阳辐照数据进行预测,提取数据中的线性分量;然后将过滤后的残差代入lstm神经网络模型,得到非线性分量的预测;最后将二者进行相加,得到最终的预测结果。仿真实验显示,组合模型比起现有的单一模型,能够有效地提高预测的精度。
在中国数字强国建设的推动下,结合计算机技术和金融知识的趋势不断增强。多因子选股模型基于长短期记忆网络(lstm),选取17个有效因子,并通过因子相关系数和|Rank IC|值筛选。模型使用多因子回归进行市值预测,通过回测验证其有效性。经过参数调优,建立最优选股模型,采用lstm算法和网格搜索优化。在2012年至2022年期间,该模型获得年均16.78%的收益率,但存在最大回撤率64.63%的风险。研究表明,lstm的多因子选股模型可获得更高收益率,承担更低交易风险。Driven by the construction of a digital powerhouse in China, the trend of combining computer technology and financial knowledge continues to strengthen. The multi factor stock selection model is based on the Long Short Term Memory Network (lstm), selecting 17 effective factors and screening them through factor correlation coefficients and |Rank IC| values. The model uses multiple factor regression for market value prediction and its effectiveness is verified through backtesting. After parameter optimization, an optimal stock selection model was established using lstm algorithm and grid search optimization. Between 2012 and 2022, the model achieved an average annual return of 16.78%, but there is a risk of a maximum drawdown rate of 64.63%. Research has shown that lstm’s multi factor stock selection model can achieve higher returns and bear lower trading risks.
暂无评论