Fig 1.
The framework of this study.
Table 1.
Data set indicator classification description.
Table 2.
Descriptive statistical tables.
Fig 2.
Statistical graph of categorical variables in data analysis.
(a) represents the mother’s education level divided into five categories, as specified in Table 1. (b) is a distribution graph of the mother’s age. (c) shows the indicators of the infants, including the distribution of infant gender, behavioral characteristics, and age.
Fig 3.
Comparing the proposed RF-MLP model with other models in the relationship between maternal indicators and infant behavioral characteristics.
When comparing the performance of the models, the AUC (Area Under the Curve) is used as the evaluation metric [38]. To ensure a fair comparison, all models are trained, validated, and tested using the same proportions of datasets.
Table 3.
The accuracy of the training, validation, and test sets of different methods.
Fig 4.
Spearman correlation coefficient heatmap.
The Spearman correlation coefficient ranges from -1 to 1 and measures the strength and direction of the non-linear relationship between two variables.
Fig 5.
The feature importance values obtained from the tree model are generally relative rather than absolute. They indicate the relative importance of features in the model for predicting infant behavioral characteristics. Positive feature importance values indicate that the feature has a positive impact on infant behavioral characteristics, while negative feature importance values indicate a negative impact. The x-axis variables represent Maternal Age, Marital Status, Educational Background, Gestation Period (weeks), Mode of Delivery, CBTS, EPDS, and HADS.
Fig 6.
Using Educational Background, CBTS, EPDS, HADS and Infant Behavioral Characteristics 370 groups of mothers and the known behavioral characteristics of infants, a RF-MLP classification model was established for prediction and a fitting plot was obtained. Among them, 1, 2, and 3 represent quiet, medium, and contradictory types, respectively.
Fig 7.
Confusion matrix.
Fig 8.
Silhouette coefficient curve.
Fig 9.
Statistical chart of sleep quality classification quantity.
The numbers of good, moderate, and poor sleep quality were 105, 216, and 57, respectively, which were normally distributed.
Fig 10.
The x-axis variables represent Maternal Age, Marital Status, Educational Background, Gestation Period (weeks), Mode of Delivery, CBTS, EPDS, and HADS. Figs (a)–(c) represent the feature value distributions for the target variables of total sleep time, number of awakenings, and sleep onset method, respectively.