Adaptive recommendation method for network resources based on improved transfer learning

To overcome the issues of high Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) values, as well as long recommendation times in traditional recommendation methods, an adaptive recommendation method for network resources based on improved transfer learning is proposed. Firstly, statistical theory is used to partition network resource data, normalise and label the data characteristics globally. Then, conventional transfer learning algorithms that are typically designed for user preferences are transformed into transfer learning algorithms for network resources. By adjusting the weight ratios of transfer learning, the learning results are enhanced. Finally, the C-CMF algorithm is applied to establish a resource adaptive recommendation process, ensuring the effectiveness of network resource recommendations through the modified processing steps. The experimental results indicate that the MAE values of this method range from 0.40 to 0.52, the RMSE values range from 0.40 to 0.55 and the average recommendation time does not exceed 2 s.