Prediction method and system for confidence interval of software cost

Abstract

The invention provides a prediction method and system for the confidence interval of software cost. The prediction method comprises the following steps of: acquiring process data and cost information of a known software project from a project management database; specifying which columns the respective process data and cost information are arranged to form a data matrix X through configuration items after data preprocessing on the basis that each row is the process data of each item of software, wherein the cost information forms a column vector Y; extracting data from an X-Y matrix, and inputting the data in a core modular module for training; training out parameters of a selected model according to input data of the X-Y matrix, and figuring out a measured value of the cost; and obtaining a final prediction interval by using the measured value of the cost and the known cost value according to a calculation method of the confidence interval. In the invention, the interval prediction method has a high hit rate of experiments, and plays an extremely important and dependable instruction role in the measurement of workload in practical application; when the prediction system is adopted, a new model, a new interval algorithm and a new model evaluation method are added for researchers; and therefore the prediction method and system are very convenient.

Description

technical field [0001] The invention relates to a method and system for predicting software cost confidence intervals based on software cost measurement, and belongs to the technical field of computer software development. Background technique [0002] The measurement of software project cost is very important for any company or organization to make investment decisions and resource coordination. Software project feasibility analysis, cost planning and schedule planning, and risk management are all based on cost determination at each stage. Therefore, the cost measurement of software projects is an important issue in the research of software project management. [0003] The current research work on software cost measurement is mainly based on the CoCoMo1 and CoCoMoII models proposed by Boehm, the originator of software cost measurement: [0004] effort = a * SIZE b * ( ...

AI Technical Summary

Problems solved by technology

In the existing measurement models, the method of predicting the confidence interval is based on the RE or MRE set, and a new model is established, which solves the problem of comparing the accuracy of the software measurement models to a certain extent, but due to the addition of the new model, it cannot reflect the measurement accuracy. The characteristics of the model itself are not conducive to analyzing the model and improving the model

Images