The exponential growth in the demands of users to access various resources during mobility has lead to the popularity of Vehicular Mobile Cloud. Vehic- ular users may access various resources on road during their journey from the cloud which acts as a service provider for them. Most of the existing propos- als on vehicular cloud use unicast sender based data forwarding, which results in an overall performance degradation with respect to the metrics such as packet delivery ratio, end-to-end delay and reliable data transmission. Most of the applications for vehicular cloud have tight upper bounds with respect to reliable transmission. Keeping in view of the above, in this paper, we for- mulate the problem of reliable data forwarding as a Bayesian Coalition Game (BCG) using Learning Automata concepts. Learning Automata (LA) are as- sumed to be the players in the game stationed on the vehicles. For taking adaptive decisions about reliable data forwarding, each player observes the moves of the other players in the game. For this purpose, a coalition game is formulated among the players of the game for taking adaptive decisions. For each action taken by a player in the game, it gets a reward or a penalty from the environment, and accordingly, it updates its action probability vector. An adaptive Learning Automata based Contention Aware Data Forwarding (LACADF) is proposed for the same. The proposed scheme is evaluated in different network scenarios with respect to parameters such as message overhead, throughput, and delay by varying the density and mobility of the vehicles. The results obtained show that the proposed scheme is better than the other conventional schemes with respect to the above metrics.