You say the results are poor for your implementation of apriori and association rules, but keep in mind that for large scale retail problems, you only need to see a fraction of per cent improvements to see significant uplift in revenues.

Association rules tell you about what customers purchase in the same transaction. This insight may not extend well to what a customer will purchase in their next transaction.

Collaborative Filtering is used in recommender systems i.e. "customers who like that, also like this." You may find it better suited to working out what a customer would like to purchase next, from a catalogue of inter-related products/services e.g. music, movies etc.

thank you for your help

Hope any Machine learning algorithms like ANN, gives you better performance