Learning to build a deep learning model based on the data of poplar tree genotypes, then to predict the phenotypes for them by building a Convolutional Neural Network(CNN) model; also to provide visualization of the prediction results.
- Five traits: "Jmax25", "Rdlight25", "Resistwp25", "WUEref" and "Av_Diameter_mm";
- Three thresholds: "1e-05", "1e-04" and "0.001".
- Firstly, we start with selecting one trait from the Poplar datasets, because we want to illustrate the general idea of building this model;
- Step by step, we read the selected data, prepare the data, build the CNN model and train the model, make prediction, and visualize the results;
- Then, we move to and work on the whole dataset, which contains 5 traits and 3 p-values. We apply the model to each traits and compare the results of the traits. Results are visualized, too;
- In addition, we apply another model to the whole dataset and visualize the new results;
- Compare the results between two different models;
- Conlude and discuss.
- Please see results for details.
Examples:
- The performance on 'Rdilight25' datasets is competently well overall;
- The smaller the p-value, the stronger stability of the model, and the better the relative comprehensive performance.
- The possible reason for the limitation of the model is that the dataset is not sufficiently large.
- Models like Linear Regression could also be applied to this problem.
- Using the model fitting measure to calculate the loss instead of accuracy.
- The objective function is to reduce the Mean Square Error (MSE) as much as possible.