In order to solve the problem that there are many construction links in construction engineering, and it is more difficult to obtain the prediction index of construction engineering cost, which leads to the decline of prediction accuracy, a prediction model of construction engineering cost based on Markov chain is designed. Through the questionnaire recovery rate, index concentration, index selection dispersion and other indicators, the principal component analysis method is used to process the index data, obtain the data feature vector, and construct the construction cost prediction index system; through the sequence law function, the transition probability matrix in the state interval is obtained, the upper and lower boundary values in different residual sequences are obtained, and the construction cost prediction results are obtained. The experimental results show that the average error of the model based on Markov chain is -0.000 25, and the average errors of the other two models are -0.045 and -0.015, respectively. The relative error of the model based on Markov chain is smaller, and the prediction accuracy is higher.