2.3. Improved collaborative filtering recommendation algorithm Based on the Internet of Things, the customer needs are recommended to help it making decisions. The input parts of the algorithm are mainly customer-item rating matrix R(u. i),customer trust model TR, customer scoring similarity threshold ε, trust threshold θ, and intersection element threshold ρ. The output part is the target customer i and the target item. The algorithm process is as follows: (1) The score similarity Sim(i, j) of the target customer i and the remaining customer j is calculated; (2) According to the similarity of the calculation, target customer i's rating similarity neighbor set in customer set U is sought, so that NNsim = {Ua |Sim(Ua, i) > ε, Ua ≠ i}; (3) Through the customer trust model, the degree of trust T (i, j) of the target customer i and the rest of the customer j is calculated; (4) The set of trusted neighbors NNtrus of the target customer i in the customer set U is found, so that NNtrust = {Ua |T (Ua, i) > θ, Ua ≠ i}; (5) if (card (NNsim ∩ NNtrust ) > ρ) Customer nearest neighbor T&S (i) = NNsim ∩ NNtrust else if (card (NNsim) > card (NNtrust)) Customer nearest neighbor T&S (i) = NNsim else Customer nearest neighbor T&S (i) = NNtrust (6) All items are networked and the target customer's prediction score for all Internet of Things items is calculated; (7) The recommended list of Internet of Things items is constructed, and all the products connected to the network are sorted, and the customer's decision is provided according to the score. End This paper fully considers the role of nearest neighbor of customer similar and nearest neighbors of customer trust, and effectively integrates them into the idea of a customer's nearest neighbor. In calculating the customer i's nearest neighbor T&S (i), the trust degree calculation of the target customer i and the rest of the customer j is performed first, and then the degree of similarity between the target customer i and the remaining customers j is calculated, and the he corresponding threshold is set. NNtrust = {T (i, j)|i, j ∈ UandT (i, j) > θ} (16) NNSim = {Sim(i, j)|i, j ∈ UandSim(i, j) > ε} (17) If the nearest neighbors of customer trust and the nearest neighbor of customer similar have the same elements, and the number of them exceeds the threshold ρ, then they belong to the nearest neighbor. Otherwise, the customer with more elements is chosen as the customer nearest neighbor. The calculation process is as follows: