Email has become one of the best ever and most efficient forms of communication. However, there are many challenges of grouping email messages. One of the challenging issues of email grouping is to identify when the clustering process accumulates accurate and sufficient information for grouping to be determined. In this work, an unsupervised machine learning technique has been developed based on unsupervised clustering method (UCM) with email vector space model (EVSM). The proposed EVSM is a novel approach as it groups emails in conjunction with UCM. Nowadays, a typical email user receives about 20-40 email messages per day. For some people depending on their job function, hundreds of messages are usual. Thus, users spend a significant part of their working time on browsing through emails. As the popularity of email communication is growing, the time spent on reading and replying emails will only increase which could lead to email overload and unstructured mailbox. As a result, there has recently been a growing interest in creating automated system to help users manage their mailbox extensively. Examples of such systems are PEA [1], MailCat [2], Re:Agent [3]. Moreover, the main tool for email management is text classification [4, 5]. EVSM is a system that implements self learning approach by The creating a space in which both email messages and queries are represented by vectors to automatically create email message groups into one or more discrete set of predefined categories. The key of using unsupervised clustering method (UCM) to cluster emails in combination with email vector space model (EVSM) will improve the performance of the email groupings being generated. Another problem that needs to be solved is the word and phrase similarity in different email messages. An email naturally has no label indicating the group it can belong to. The label can only be set by humans for accuracy, which needs heavy manual labor. This challenging issue is especially significant for email users who need to be labeling email messages frequently in order to manage and organize their inbox. 2. Email Vector Space Model (EVSM) This work developed a novel approach that uses both the structure and content of email messages for email grouping. Since emails composes of a structure in the form of headers and the message body, they can be represented using feature selection and the relationships between various terms (e.g., the occurrence of a term in the subject or body of the message etc) can be represented in the form of a cluster. EVSM has been developed as a viable technique for pattern extraction, frequent words and phrase selections. This approach is based on the premise that representative – common and recurring – structures/patterns/features can be extracted from a pre-grouped email dataset and the same can be used effectively for grouping incoming emails. It suggest that an email dataset consists of representative emails or knowledge sets which contains typically a set of keywords representing a specific meaning, and the structure and content of these representative emails or knowledge sets can be extracted by adapting unsupervised email vector space model techniques to work with the developed UCM. The work also, hypothesise that the notion of clustering is critical for this to work as it helps in grouping similar structures within emails instead of looking for exact/identical matches that may be difficult to find in the email messages.