在纵向时间轴上设定滑动窗口来动态选择商品合适的同期历史记录,并在所选时间序列中巧妙地将滑动窗口与最佳(Pearson Correlation)拟合公式和均值生成函数相结合。提高了算法的灵活性和预测值的精度。该方法的主要优点包括过滤异常数据记录,避免因其产生负面影响,并能得到清晰的商品销量变化趋势。实验证明,该方法无论从时间复杂度还是预测准确度来说都是可行的。
Sliding window on vertical time axis is first used for selecting historical data similar to the predicted goods. Historical data in the sliding window is processed using both Pearson Correlation and average value generation function so as to improve the flexibility of the algorithm and the prediction aecuracy. The remarkable advantages of the approach include filtering exceptional data, and generating the trends of sales of goods. Some experiments demonstrate that this algorithm is efficiency at both the time complexity and promising accuracy.