Query Processing on Big Data has received significant attention in the literature. Many Big Data sources generate unstructured data. Big Data is characterized by extreme velocity of data generation and very large data volumes, and due to which, structuring of data becomes impractical. Hence, Big Data queries need to be executed on unstructured data, which results in extreme computational costs. Hence, many Approximate Query Processing Techniques (AQPTs) have been presented in the literature which provide approximate query results using sampled data, and thus achieve noticeable computational efficiency. Recently in the literature, AQPT is presented for the approximate execution of simple non-join aggregate queries. This presented AQPT achieves a predefined estimation error, and exhibits noticeable computational efficiency, however, this AQPT only addresses simple non-join aggregate queries, and does not address join-aggregate queries involving join of multiple relations. To address this open issue, in this paper, AQPT is presented for the approximate execution of join-aggregate queries which involve join of multiple relations. The proposed AQPT achieves pre-defined estimation error. Empirical analysis study of the proposed AQPT along with the contemporary technique is outlined. In this outlined empirical analysis study, the proposed AQPT significantly outperforms the contemporary technique in-terms of estimation accuracy and query execution latency.