Predictive Modelling of Different Stress Treatments for Physiological and Nutritional Parameters of Ocimum sanctum

Abstract

An increasing demand for fast and healthy production crops has led to investigating the current trends in agriculture pertaining to their quality production. One of the main factors impacting this is environmental stress. Even though plants have developed defences to combat these stresses, not all plants can tolerate the impact induced by abiotic stresses [1]. Due to this, the overall physiological and nutritional parameters are greatly compromised.

The study focuses on the methodologies involved and results obtained from growth and collection of plant parametric data which is further processed and analyzed by predictive models. Ocimum sanctum, commonly known as Tulsi [2,3], is introduced to various abiotic stresses in order to study its physiological and nutritional variations observed and compared to normal conditions [4]. These sample plants were subjected to various abiotic stresses – drought [5], jasmonic acid [6], potassium nitrate [7], hydrogen peroxide [8], at varied concentrations, and their effects on certain parameters - plant height, relative water capacity, protein and chlorophyll content and photosynthetic rate were subsequently recorded over a period of 50 days. This data set was used to train certain machine learning models for prediction of specific treatments and parameters.

The k-nearest neighbor classification model [9] was used to predict the treatments required for any given parameter. A comparative study between Linear Regression model and Non-Linear Regression model [10,11] with the same input data and a common output of predicting one parameter from the other known parameter was also performed and were able to infer that Non-Linear Regression model performed well and had a better accuracy (approximately 85%) compared to the former for the given amount of data.

Methodology

The methodology can be briefly divided into three different phases;

2.1 Experiment set-up

The first phase which served as the foundation, is the experimental set-up. It was created after formulating the working principle and theory. The experiment was set on 20 potted plants of Ocimum sanctum for each stress condition. After 15 days of planting, stress (drought, 50 mM KNO3, 50 µM and 100 µM jasmonic acid, 2% H2O2) was subjected to the plants.

2.2 Measuring the parameters

The second phase was to measure the different parameters such as plant height, relative water capacity, protein content, chlorophyll content and photosynthesis efficiency for each plant. The measurements were taken from day 2 after stress treatment to day 50 with the intervals of initially 2 days and then 5days.

2.3 Statistical analysis, prediction using classification and regression models

The third section, which exhibits the novelty of this paper, focuses on the utilization of the data obtained from the second phase, with the goal to observe the trends in the data, and to subsequently make a prediction of stress treatment and equivalent parameters. Each parameter across all the stresses were represented through graphical analysis. The k-NN classification model was used to predict the class of treatment while linear and non-linear regression models were used to predict the corresponding unknown parameter. The conclusions and inferences drawn from the phases undertaken as part of this study will be used in defining the future scope.