The growth of the bid data set has become an integral part of business enterprises. Still, for responding to bids, predicting the making decision is considered as the most important. So, many techniques and algorithms become a crucial solution. In this context, artificial intelligent techniques for prediction are applied to aid in decision making. In this study, we consider machine learning algorithms as a prediction problem based on the random forest, where the goal is to enhance the performance of this algorithm coupled in a Big Data approach. To do this, instead of randomly dividing the characteristics to each node of each tree, we consider using a dynamic characteristic selection to each tree. This can help to introduce more diversity into trees and potentially improve the performance of random forest to be a dynamic forest algorithm. Thus, we compare it without dynamic characteristics (static) in terms of ROC curve, precision-recall and accuracy. The findings obtained demonstrate that the random forest model with dynamic feature selection is more effective than the static feature selection in predicting the final decision, with a 97 % accuracy rate for dynamic features and a 94 % accuracy rate for static features.

中文翻译：

---

人工智能集成，用于扩展大数据以进行决策


bid 数据集的增长已成为企业不可或缺的一部分。尽管如此，对于响应出价，预测做出的决定被认为是最重要的。因此，许多技术和算法成为至关重要的解决方案。在这种情况下，人工智能预测技术被应用于辅助决策。在这项研究中，我们将机器学习算法视为基于随机森林的预测问题，其目标是提高该算法在大数据方法中耦合的性能。为此，我们考虑对每棵树使用动态特征选择，而不是将特征随机划分到每棵树的每个节点。这有助于为树引入更多多样性，并可能提高随机森林的性能，使其成为动态森林算法。因此，我们在 ROC 曲线、精度-召回率和准确性方面比较了没有动态特性 （static） 的它。获得的结果表明，具有动态特征选择的随机森林模型在预测最终决策方面比静态特征选择更有效，动态特征的准确率为 97%，静态特征的准确率为 94%。