With the important changes in the era of the big data and digital transformation, data modeling is essential in unlocking the value of data and assisting in decision optimization. But the conventional data modeling is plagued with a myriad of issues. The feature extraction is based on human experience and therefore, it is low in efficiency, and it can easily overlook some important information; modeling iteration is time-consuming, the generalization power is low and it can hardly work with multi-domain and the high-dimensional data hence this is where the computer algorithms come in to help remove these bottlenecks. Thus, the paper initially discusses the current research position of computer algorithms and the modeling of this data, detailing its drawbacks; next, it creates the general technical framework, developing solutions based on three dimensions, that is, the optimization of the feature mining algorithm, the adaptation of the modeling process, and the creation of an efficiency evaluation system to enhance the accuracy of feature mining, and the efficiency of the modeling. Lastly, simulations are performed with the datasets of the fields of financial risk control, image recognition, and text classification to prove the possibility and high effectiveness of the offered solution. The proposed algorithm has a feature extraction accuracy of 95%, which is 13 and 9 percentage points higher than support vector machines and random forests, respectively. It can effectively adapt to complex scenarios and analyze application value.
Tian et al. (Thu,) studied this question.