| Red carbon dots are a novel carbon-based nanomaterial with superior optical properties.This material is capable of emitting red fluorescence,and has broad application prospects in various fields,such as biological detection,cellular imaging,targeted labeling,environmental monitoring,and many others.However,as the luminescence mechanism of red carbon dots remains unclear at present,the majority of their acquisition methods rely on traditional trial-and-error approaches.The trial and error method requires designing multiple sets of experiments in advance and then synthesizing them one by one,aiming to increase the probability of obtaining red carbon dots through a large number of experiments;This leads to low synthesis efficiency and high acquisition costs.Moreover,the fluorescence intensity of red carbon dots is severely affected by the pH value,thus monitoring the pH environment is a crucial step to exploit its fluorescent properties.However,conventional methods for pH detection require the use of expensive materials and instruments,and the procedures are cumbersome and time-consuming,which further increases the application cost of red carbon dots.Therefore,optimizing the synthesis method of red carbon dots and developing more efficient and low-cost pH detection methods are of great significance for the application research of red carbon dots.In order to reduce the difficulty of obtaining red carbon dots,Machine learning was used to solve issues in obtaining red carbon dots,resulting in established solutions.Work contents as follows:1.To address the issue of low efficiency in the synthesis of red carbon dots,an efficient synthesis method for red carbon dots was established.This method first constructed a dataset regarding the synthesis conditions of red carbon dots based on the knowledge in the field of carbon dots.Subsequently,a weighted feature selection strategy based on SHAP importance was established,and the optimal feature subset that affects the color of carbon dots was obtained.The effectiveness of this strategy was verified through comparative experiments.Finally,the traditional Stacking fusion strategy was optimized through the data transformation layer and the base learner selection layer,and a carbon dot color prediction model was established.In addition,multiple sets of carbon dot synthesis experiments were designed and implemented independently,and the experimental results were compared with the predicted results of the model,which verified that the carbon dot color prediction model has a high accuracy.Subsequently,a series of detection experiments and cell imaging experiments were performed on the synthesized red carbon dots.The experimental results confirmed that the pH environment greatly affected the fluorescence intensity of the red carbon dots,indicating that the red carbon dots have good performance as a biological marker.2.To address issues such as low detection efficiency and high detection costs in pH detection,an efficient detection method for pH environment was established.This method first uses inexpensive and easy-to-operate pH test strips for preliminary pH value detection,and then utilizes digital image technology to construct artificial features from the image data of the test strips,obtaining structured color feature datasets.Secondly,after comprehensively considering the importance and correlation of features,a correlation feature selection strategy based on SHAP importance was established and applied to eliminate a large number of redundant features.The effectiveness of this strategy was verified through multiple comparative experiments.Lastly,a prediction model for pH value was established based on the Stacking fusion strategy,with a prediction error of only 0.039.3.From the practical point of view,a color prediction system for carbon dots and a pH detection system were developed by utilizing frameworks such as Django and Bootstrap,and the integration and implementation of the two methods were carried out. |
PDF Full Text Request