Sentimental analysis or Opinion mining is used in analyzing the important opinion from the reviews generated by the users. The main objective of Sentiment analysis is classification of opinions/sentiments. It classifies the given text in three levels: document level, sentence level, and entity/aspect level. The accumulation of vast and unstructured opinions on many domains has been making information gaining difficult. Opinion mining is the opening technique towards tackling this problem. In this research work, an Opinion Mining model is built for classifying Amharic opinionated text into positive, neutral and negative. The experiments are conducted using 787 Amharic opinionated texts collected from Amhara broadcasting agency Facebook page and YouTube cannel users’ sport and business news comments. We use term frequency invers document frequency feature extraction method and implement supervised classifiers from the Natural Language Toolkit in Support Vector Machine algorithm. The design and development of this thesis work is based on the machine learning approach. The experiment results 83.24% Accuracy in Support Vector Machine algorithm, Generally, the results are encouraging despite the morphological challenge in Amharic, the data cleanness and small size of data. Negations and valence shifters will be considered as a feature in ML approach because their presence in the sentence can result in changing the sentiment of the whole comment like “ጥሩ - nice” implying positive sentiment if preceded by “ጥሩ አይደለም– not good” would then imply negative sentiment.