News summarization is a process of distilling the most important information from a news document in a precise way. For the advancement of Internet nowadays almost all of the Bangla newspapers have their online versions, and people of this era like to read newspaper from website using Internet. But large amount of electronic news content is a burden for human to come out with valuable information. For mitigating this pain point, this paper proposes an automatic method to summarize Bangla news document. In this proposed approach, graph based sentence scoring feature is introduced for the first time for Bangla news document summarization. After analyzing vast amount of Bangla news document 12 sentence scoring features have been introduced for calculating score of a sentence. An improved summary generation method has also been proposed which remove the redundant information from summary. The result is evaluated using a standard summary evaluation tool called ROUGE, and found proposed method outperforms all existing methods used in Bangla news summarization.