Construction Of Machine Learning Models For Predicting Intracellular Carbohydrates Accumulation Based On Algae Photosynthetic Parameters

Microalgae converts the carbon dioxide into lipids and carbohydrates through effective photosynthesis and self-metabolism;then,the microalgae stores them in cells,which can be utilized as raw materials for biomass ethanol production.However,during the process of traditional methods determining the intracellular carbohydrates contents of microalgae,there will be limitations including secondary chemical pollution,cumbersome steps and being uncertain of the carbohydrates content of microalgae cells in real time.Absolutely,the production of carbohydrates is closely related to photosynthesis;meanwhile,artificial intelligence rapidly develops in the statistical processing of biological information.In recent years,with the vigorous development of deep learning in artificial intelligence,a new opportunity provides new opportunities for efficient processing of complex information.On one hand,the accumulation of carbohydrate in the microalgae is closely related to photosynthesis.On the other hand,the efficiency of photosynthesis can be rapidly evaluated by the fluorescence parameters of the leaf germin,and the chlorophyll fluorescence parameters are the link for depth learning method being constructed with the advantage of unintrusion.The machine learning test model of secondary chemical contamination is of great significance to the innovation of the detection methods in the field.Within the essay,the first part conducted the experiments that under different temperature gradients,a variety of environmental indicators and physiologies of the microalgae was monitored,which involved microalgae biomass,consumption of nitrogen and phosphorus nutrients,carbon fixation rates,chlorophyll fluorescence parameters,proteins,photosynthetic pigments,lipids and carbohydrates contents.Through temperature optimization,the optimum growth temperature of this strain is about 30°C,and its biomass can reach 1211 mg/L,the highest carbon capacity in the second day.The maximum carbohydrate content was obtained on the 6th day of the experiment,is 75.16%.As the experiment is carried on,the chlorophyll fluorescence parameters are all better in front of the source of the nitrogen source,such as the increase in photon yield,speed of electron transfer rate.With the depletion of nutrients,the electron transfer of proton quinone and the photon yield decreased.The yield decreased,and the chlorophyll fluorescence parameters were harvested.At the same time,the content of the intracellular carbohydrate content of microalgae has also increased first,and there is a tendency to decline.Low temperature causes the growth rate and metabolic activity of microalgae to inhibit,but the content and components of the carbohydrate have not changed much.At the same time,it can be determined that the influence of temperature difference within a certain range on carbohydrates accumulation is less than that of nutrient deficiency.The second part of this paper organizes the experimental data of the foregoing section and collects data sets from the existing research literature.After interpolation and normalization of the data,five machine learning models such as multi-linear regression,artificial neural network,Xg Boost,random forest regression,support vector machine regression,they are constructed to predict the carbohydrates of chlorella in percentage contents.Then,rooted root error and decision coefficient R~2 was both utilized to evaluate the fitting effects of these models.The construct of the support vector machine was selected as a nuclear function,which had a significant advantage for small sample processing and was not easily affected by strong impact points.Therefore,the root mean square error of the support vector machine regression model is 0.0745,and the decision coefficient R~2 is 0.9254,and the best predicted fit effect is achieved.In addition,among the five models,the multi-linear regression model had better control variable interpretation;and the artificial neural network model was not interpretable,but the shared foundation was laid into the future information input and storage.The random forest regression model significantly reduced the complexity of the model due to its regularization,which in turn effectively preventd the prefraction.This topic finally established a machine learning model predicted by chipocyte carbohydrate content predicted in chlorocyanocytes in chloride fluorescence parameters.Since the input parameters of the machine learning model have the characteristics of fast instant,no invasive,these machine learning models are of great significance in evaluating microalgae growth,and determining the yield of microalgae carbohydrates,determining the harvest time of microalgae.