Innovative Mechanism of Rural Finance: Risk Assessment Methods and Impact Factors of Agricultural Loans Based on Personal Emotion and Artificial Intelligence

Table 2

Accuracy of variable-length farmers’ speech in agricultural loan review emotion recognition model based on CNN + Bi-LSTM.

Model	Spectrogram input form	WA (%)	UA (%)
Three convolutional layers + Bi-LSTM	Fixed-length input	84.26	77.01
Three convolutional layers + Bi-LSTM	No padding input	61.55	45.00
Three convolutional layers + Bi-LSTM	Batch padding	85.61	80.73
Three convolutional layers + Bi-LSTM	Maximum length padding	84.99	79.43