Cell Carcinosis Prediction using Linear Regression with Nuclear Statistics

Cell carcinosis prediction mainly focuses on detecting cancer cells based on specific physical characteristics, which help to diagnosis, early prevention, and treatment. The prediction of cell cancer depends on many factors, such as the diameter, area, smoothness, and tightness of the nucleus. This paper uses model construction and linear regression to predict cell cancer with nuclear statistics to assist doctors in a cancer diagnosis. Firstly, this paper performs a rough parameter selection to filter out those with obvious correlations. In order to obtain parameters and to simplify the calculation, this paper eliminates redundant parameters. Then, the F-test is used to judge the significance of parameters and select the most useful parameters for model construction. After that, the variance inflation factor (VIF) is used to test whether the multicollinearity problem is solved, Difference in fits (DFFITS) is used to detect outliers. The experimental result shows calculated mean VIF for this model is 2.46, the multicollinearity problem is solved. All DFFITS values are lower than 1. There are no outliers. Our analysis shows that a function that indicates the accurate probability of cell carcinosis, and when we plug in specific cell data, we can obtain the degree of cancerous cell.