In recent years, the popularity of social media networks has attracted the attention of researchers, government agencies, politicians and business world alike, as a powerful platform to explore real-time trends. The data generated by these networks offers an opportunity to investigate people's behaviors and activities, but the high velocity and low quality of this data poses some unique challenges. Twitter, an example of social media networks, is particularly popular for this purpose due to its easily accessible API that is open to use for research purposes. Different techniques can be used to analyze patterns from available data. One of such techniques for extracting subjective information from any text such as opinions on various topics is Sentiment Polarity Classification, which quantifies emotions embedded in texts and classifies them as positive, negative or neutral. The focus of this paper is on preparing and analyzing real-time twitter streams to detect real-time trends on a particular topic using Sentiment Polarity Classification. We have used StreamSensing approach and have performed a supervised machine learning on real-time high velocity data using Apache Spark micro-batching technology to classify the opinions and feelings of people in real-time. Appropriate experiments for processing high rate of incoming streams have been carefully designed and conducted on live twitter data. The outcomes of these experiments were analyzed and presented. The findings of this paper fell into two perspectives: theoretical and practical. The theoretical perspective is seen in testing and confirming the validity of StreamSensing approach as well as the introduction of a sentimental polarity algorithm, while practically; this approach can be employed to perform trend analyses on any real-time streams related to live events.