基於機器學習建立新聞分類模型
－應用於線上新聞瀏覽平台

在這資訊爆炸的時代，隨著科技的進步，每天出現新的新聞量大增，但這些訊息分散在各大新聞平台上，而某些新聞平台可能因為政治立場不同、作者喜好…等等因素，造成各新聞平台公布出的新聞內容有所差異。只單獨接收一個新聞平台的新聞容易被內容帶風向，而失去判斷的能力，因此接收來自各個不同平台的新聞就顯得非常的重要。假如有一個新聞平台能收取各個不同平台的新聞供使用者來瀏覽，可省去很多搜尋新聞的時間。而各大新聞平台的新聞分類都不相同，如果使用人工的方法分類新聞會浪費很多人力、時間以及成本，顯然是不切實際的作法。此外，根據不同的分類者，同一篇新聞可能得出的分類結果也不相同，因此，建立機器學習的模型，交給模型做自動分類就至關重要。藉由收集各類不同的新聞文章，使用機器學習(Machine Learning)來建立模型，讓機器做自動分類的功能，能省去很多人工分類的時間。研究結果顯示使用建立好的模型做出分類，已達成97%的正確率。

關鍵字

機器學習；新聞分類； SGDClassifier

並列摘要

In this information booming era, with the progress of technology, there are huge amount of news pop up everyday. Yet, these messages are dispersed to each news platform. However, because of different political views, the preference of the authors and other factors, it may cause the news contents announced by each news platform to have differences. The audiences may easily be spin-controlled if they only absorb news from one news platform, thus, their abilities to judge may be lost. Hence, it is apparently very important to receive news from each different platform.Assumed that there’s a news platform that collected news from every different platform for the readers to browse, it can save a lot of searching times for them. If using manual to classify news, a lot of manpower, times and costs will be wasted, so obviously, it is an unrealistic way to do. In addition, for the same news, it may have different classifying results due to different classifiers. Therefore, to build a machine learning model to let the model classify automatically will be extremely crucial. All in all, through gathering each type of news articles to use machine learning to build a model and let the machine to have a function of classifying automatically will save a lot of time for the manpower to classify. The result of the research indicating that, adopting the built model to classify has already reached 97% of the accuracy.