This paper presents a new Monte Carlo algorithm for calculating the density of states (g(E)) with high accuracy and reduced computational effort. The method involves performing independent random walks in different energy ranges, which are then combined to produce a flat histogram in energy space. This allows direct access to free energy and entropy, and is efficient for studying both first-order and second-order phase transitions. The algorithm is particularly useful for complex systems with rough energy landscapes. The algorithm works by estimating g(E) through a random walk where the probability of transitioning between energy levels is proportional to the reciprocal of g(E). Initially, g(E) is set to 1 for all energies, and the density of states is updated as the random walk progresses. The modification factor f is adjusted over time, starting from a high value (e.g., e) and decreasing until it approaches 1. This ensures the histogram becomes flat, indicating convergence to the true g(E). The method is tested on the 2D Ising model and the 2D ten-state Potts model, demonstrating high accuracy and efficiency. For the Ising model, the algorithm achieves an average error of 0.035% on a 32x32 lattice with 7x10^5 sweeps. The algorithm also allows the calculation of thermodynamic quantities like Gibbs free energy, entropy, and specific heat with high accuracy. For the Ising model on a 256x256 lattice, the specific heat error is reduced to below 0.7%. The algorithm is efficient for large systems and avoids the need for multiple simulations at different temperatures. It is also effective in overcoming tunneling barriers between coexisting phases, as demonstrated by its application to the 2D ten-state Potts model. The method is applicable to various systems with complex energy landscapes, including spin glasses and protein folding problems. The algorithm's efficiency and accuracy make it a valuable tool for studying complex systems in condensed matter physics.This paper presents a new Monte Carlo algorithm for calculating the density of states (g(E)) with high accuracy and reduced computational effort. The method involves performing independent random walks in different energy ranges, which are then combined to produce a flat histogram in energy space. This allows direct access to free energy and entropy, and is efficient for studying both first-order and second-order phase transitions. The algorithm is particularly useful for complex systems with rough energy landscapes. The algorithm works by estimating g(E) through a random walk where the probability of transitioning between energy levels is proportional to the reciprocal of g(E). Initially, g(E) is set to 1 for all energies, and the density of states is updated as the random walk progresses. The modification factor f is adjusted over time, starting from a high value (e.g., e) and decreasing until it approaches 1. This ensures the histogram becomes flat, indicating convergence to the true g(E). The method is tested on the 2D Ising model and the 2D ten-state Potts model, demonstrating high accuracy and efficiency. For the Ising model, the algorithm achieves an average error of 0.035% on a 32x32 lattice with 7x10^5 sweeps. The algorithm also allows the calculation of thermodynamic quantities like Gibbs free energy, entropy, and specific heat with high accuracy. For the Ising model on a 256x256 lattice, the specific heat error is reduced to below 0.7%. The algorithm is efficient for large systems and avoids the need for multiple simulations at different temperatures. It is also effective in overcoming tunneling barriers between coexisting phases, as demonstrated by its application to the 2D ten-state Potts model. The method is applicable to various systems with complex energy landscapes, including spin glasses and protein folding problems. The algorithm's efficiency and accuracy make it a valuable tool for studying complex systems in condensed matter physics.