近年來，室內外定位偵測發展蓬勃，已經成為現今不可或缺的技術之一，並具有許多潛在的應用。舉例來說，可以透過定位使用者是否有從室外環境進入到室內環境， 以便協助使用者手機自動關閉GPS功能來節省能源，或是讓使用者手機自動切換靜音模式。本論文利用tree-based learning 演算法 (i.e., decision trees, boosting, bagging, and random forest) 來分類室內與室外定位的資料。並使用10-fold cross validation來驗證分類結果，以避免分類結果有overfitting的問題發生。最後比較各個tree-based learning 演算法分類結果的性能，並找出其中最適合分類室內外定位資料的tree-based learning演算法。在我們實驗中，雖然分類結果普遍偏高，但其中以boosting的演算法為最佳。Boosting在室內外定位資料分析有高達99%以上的正確率。;The indoor/outdoor detection for wireless device has many potential applications. For instance, when a device is detected to enter the indoor environment, it can turn off the GPS chip to save energy. In this thesis, the indoor/outdoor detection is treated as the supervised learning problem. The tree-based learning algorithms (i.e., decision trees, boosting, bagging, and random forest) are built from the training dataset and used to classify test dataset. In addition, the 10-fold cross over is used with the algorithms to mitigate the issue of overfitting. The performance of each algorithm are compared to identify the algorithm most appropriate for the indoor/outdoor detection. Although the final performance of each algorithm seems to be high, boosting provides the best accuracy (99%) for indoor/outdoor detection.
