This paper reviews the basic principles of the random walk, and proposes two modifications for the technology, related with the random influencing factor and the widening channel methods. First, in the calculation of the transition probability, the Monto Carlo factor is added to increase the stochastic extent and to make the model more deterministic. Second, the circling filling approach is used to reconstruct the width of the channel along the channel mainstream, so that the channel will not vary sharply in the high curvature points. The real data from the eastern oilfield is used to test the modified algorithm. It is shown that it is similar to drawing picture by a geologist and the results agree with the real geological phenomenon, with a coincidence rate of 76.5%. Furthermore, the results of the modified algorithm using the sparsing testing agree with the results before the sparsing, with a rate of matching of 77.3%. Finally, some problems in using the random walk are analyzed and related solving schemes are proposed.