Generalist and specialist predators are successfully used in biocontrol programs in greenhouse vegetable crops, like tomato. A greenhouse ecosystem is rather simple and provides an excellent opportunity for developing predator–prey decision models. Three systems were selected: (1) the generalist predatory bug Macrolophus pygmaeus and the greenhouse whitefly Trialeurodes vaporariorum, (2) the generalist predatory bug Nesidiocoris tenuis and the tobacco whitefly Bemisia tabaci and (3) the specialist predatory mite Phytoseiulus persimilis and the spider mite Tetranychus urticae. The study is based on an extensive field dataset. No complex mathematical predator–prey models were developed. A binomial variable was given the value of “0” for the period when the pest was not under control. As soon as the population declined after the peak density, this variable was given a value of “1”. The relationship between the densities of the prey and the predator was checked using a logistic regression model. The validated models do not calculate future pest densities but rather predict when pest control should be initiated, based on the number of pests and predators present at a certain time. Numerical simulation of the prey isoclines showed an L-shaped curve for the generalist predators and a linear curve for the specialist predators. Our simple, empirical modelling approach provides satisfactory models for biocontrol purposes. When combined with a standardized monitoring protocol, these models can be implemented in decision tools. In the future, more data will allow a machine learning approach, in which additional parameters like temperature, humidity, and time can be included.

通才和专家捕食者已成功用于温室蔬菜作物（如番茄）的生物防治计划中。温室生态系统非常简单，为开发捕食者－猎物决策模型提供了极好的机会。选择了三个系统：（1）通才的捕食性小虫Macrolophus pygmaeus和温室粉虱Trialeurodes vaporariorum，（2）通才的捕食性小虫Nesidiocoris tenuis和烟草烟粉虱Bemisia tabaci，以及（3）专业的捕食性螨Phytoseiulus persimilis和蜘蛛螨螨。该研究基于广泛的现场数据集。没有开发复杂的数学捕食-被捕食模型。在有害生物不受控制期间，将二项式变量的值设置为“ 0”。一旦人口在峰值密度之后下降，该变量的值将为“ 1”。使用逻辑回归模型检查猎物和捕食者的密度之间的关系。经过验证的模型不会计算未来的有害生物密度，而是根据特定时间存在的有害生物和捕食者的数量预测何时应开始控制有害生物。猎物等高线的数值模拟显示，通配捕食者呈L形曲线，而专业捕食者呈线性曲线。我们简单的经验建模方法可为生物防治提供令人满意的模型。与标准的监视协议结合使用时，可以在决策工具中实现这些模型。将来，更多的数据将允许机器学习方法，其中可以包括温度，湿度和时间等其他参数。