The invention discloses a safe and efficient sensor
data prediction and search method, and belongs to the field of
Internet of Things. The method comprises the steps of region division,
data acquisition, model construction, joint reasoning, search
initiation and search response. Wherein the region division comprises the step of dividing the terminal sensor equipment into a fixed management range according to the geographic position of the terminal sensor equipment, and uniformly managing the terminal sensor equipment by a belonging
edge server. The data collection comprises the steps that the sensor collects periodic data or non-periodic data according to the
function type of the sensor and uploads the periodic data or the non-periodic data to the
edge server. And model construction: constructing a
federated learning framework based on edge cloud
collaboration, and initializing to construct a sensor
data prediction model based on T-LSTM. The joint reasoning comprises the steps that the edge servers and the
cloud server cooperatively
train prediction models, and the trained prediction models are deployed to the edge servers. Search
initiation comprises the step that a user initiates a search request to the
edge server through the intelligent
client. The search response comprises the steps that the edge
server recognizes a user request and executes search operation, and required sensor
data information is fed back to the user. The method can effectively improve the search efficiency of the sensor data and protect the data privacy at the same time.