In unattended sensor networks, a major challenge in data survivability. The data collected by sensors can be lost due to various reasons like sensor node failures, attacks etc. In our earlier work, a data survivability solution based on mobile sink with increased focus on security and data compression was proposed. But the network structure lacked efficiency to enhance data survivability in terms of data backup at optimal locations and collection of them before any risks. This work addresses this problem and proposes a improvised mobile sink based data survivability scheme. The data collection points are optimized with aim of enhancing data survivability using graph theory concepts. Through simulation analysis the proposed solution is found to increase data survival probability by atleast 3% compared to existing works. The proposed solution was also found to be energy efficient and increased the life time of sensor network by atleast 11% compared to existing works.