Odor is a type of sensory pollution that can stimulate the human sense of smell when it occurs, causing discomfort and making it difficult to create a pleasant environment. For this reason, there is a high possibility of complaints regarding odors if odors occur in pigsties near residential properties, and the number of such complaints is also increasing. In addition, odors emanating from pigsties around military installations can cause physical and psychological harm, not only to the soldiers living in these type of facilities but also to the families belonging to military personnel living there as well. Because the concentration of odors varies due to diverse factors such as temperature, humidity, wind direction, wind speed, and interaction between causative materials, predicting odors based on only one factor is not proper or appropriate. Therefore, in this work, we sought to construct models that are based on several regression techniques of machine learning using data collected in field. And we selected and utilized the model that has the highest-accuracy in order to notify and warn residents of odors in advance. In this work, 3672 data items were used to train and test the model. The several machine learning algorithms to build the models are polynomial regression, ridge regression, K-nearest neighbor regression (KNN Regression), and random forest. Comparing the performance of models based on each algorithm, the study found that KNN Regression was the most suitable model, and the result obtained from KNN regression was significant.

이전에 학습한 자료에 대하여 시험을 쳐 보는 것은 강력한 학습 이벤트로 작용할 수 있다. 이를 시험 효과라고 하는데, 본 연구에서는 시험 효과 검정을 위한 중세국어-현대국어 단어쌍을 개발하 여 사용하였다. 본 연구에서는 이를 활용하여 초기 테스트, 단서 회상 테스트, 단기 파지 간격에 따른 집단 간의 차이를 살펴보았다. 그 결과, 초기 테스트 집단이 매우 유의미한 결과를 나타내었 다. 이것은 학습 후 빠른 시간 안에 초기 테스트가 필요함을 강력히 지지한다. 그리고 단서 회상 테스트가 주어질 경우 더욱 그 효과가 뛰어났다. 그러므로 실제 교실 수업에서도 학습 후 테스트나 퀴즈의 형태를 통해 학습자의 학습을 도울 수 있으며, 형성평가 등을 적극 활용한 교수․학습을 설계할 것을 제안한다.