The phrase "data imbalance" is correct and usable in written English.
It can be used in contexts related to statistics, machine learning, or data analysis, where there is an unequal distribution of data across different categories or classes.
Example: "The model's performance was affected by data imbalance, leading to biased predictions."
Alternatives: "data disparity" or "data skewness".
The F1 score is weighted by the class frequency to account for the data imbalance.
Considering the extreme data imbalance regarding wildfire ignitions and nonignitions, a cost-sensitivity analysis is usually used in the models.
Thus the data imbalance will cause the insufficient study on properties of default sample, which has bad effect on the classification accuracy of both SVM and BP model.
Because it is easier to obtain good credit enterprises' information than bad credit enterprises, so there is data imbalance in the training sample set.
However, because of the high intraclass variance in real world image classification, the One vs. One method suffers from the same high data imbalance problem.
Careful processing of data samples is needed, however, to resolve issues of data imbalance and to avoid potentially misleading results due to the overwhelmingly large number of nonignition samples.
Due to the special characteristics of the defect prediction data (imbalanced, inconsistency, redundancy) not all classification algorithms are capable of dealing with this task conveniently.
The data imbalances both inside each class and between classes are addressed by dedicating variable weights to kernels as discussed in section 1.2.
This method resolves a great deal of the poor model classification performance that results from data imbalances (Del Río et al. 2014).
There is a necessity for a thorough design at the implementation level for current algorithms In other words, an effort for the design and development of robust methodologies to address Big Data imbalanced problems has to be made.
To maximize performance, samples from the positive (minority) class were weighted to compensate for data imbalances, and model parameters were tuned in ten-fold cross-validation such that the minimum number of trees exceeded 1,000 [22], [23].
