Читать книгу Data Mining and Machine Learning Applications - Группа авторов - Страница 54

2.4 Results

Оглавление

The method is developed to predict how much a development/decrease adjustment will occur based on input factors such as time and inner situation. In the time leading up to measurement, the planning set was standardized, meaning that all highlights are rescaled to zero mean and unit-fluctuation dispersions. At that point, the dataset is cared for in an L1-punished strategic relapse classifier, which will streamline the cost capacity to predict the response of residents in a particular situation. As the portion scale is normalized, the prepared straight model coefficient may show the overall meaning of the compared element. For example, Figure 2.11 Indicates the importance of each trigger factor for tenant No. 1, with the model being 86% inter-approved.

It could be seen that the less instructive highlights for this inhabitant were sifted through with zero coefficients, while the remaining shows the indoor CO2 focus and dampness are the most significant inspirational drivers for this tenant to change the ventilation stream rate. By this methodology, the primary driver for inhabitant No. 1 to alter ventilation flowrate is distinguished.


Figure 2.11 Highlight significance yield.

Data Mining and Machine Learning Applications

Подняться наверх