The invention establishes an intelligent evaluation algorithm system based on collateral big data. A value evaluation process uses two networks. A first network is used for determining an evaluation algorithm the most suitable for the collateral and increasing the evaluation weight of the algorithm. The second network is used for collecting user feedback information. Whatever correct evaluation, deviated evaluation, or erroneous evaluation, on the one hand, a computer automatically analyzes and records the evaluations, and on the other hand, a worker participates to help the computer correct the deviation or the error, which is similar to a deviation or error memory stick, and all evaluations are fed back to the first network. Two networks continuously and cyclically operate, learn the correctness and the wrongness, form experience after long-time accumulation, perform continuous improvement, and increasingly improve the accuracy rate in order to achieve automatic batch precise evaluation, and achieve risk analysis based on the precise evaluation.