Data Science
Budget: $250 – $750 USD
Topic: Predicting the final properties of new materials (composite materials).
Description:
Composite materials are artificially created materials, consisting of several others with a clear boundary between them. Composites have properties that are not observed in the individual components. In this case, composites are a monolithic material, i.e., the components of the material are inseparable from each other without destroying the structure as a whole.At the input there is data on the initial properties of the components of composite materials (the amount of binder, filler, curing temperature, etc.). At the output, it is necessary to predict a number of final properties of the obtained composite materials.
Dataset with properties of composites. Join do by index join type INNER
https://drive.google.com/file/d/1B1s5gBlvgU81H9GGolLQVw_SOi-vyNf2/view?usp=sharing
Required:
1) To study the theoretical foundations and methods for solving the problem.
2) Conduct an exploratory analysis of the proposed data. It is necessary to draw histograms of the distribution of each of the variables, box and whisker plots, pairwise scatter plots. It is also necessary to obtain an average, median value for each column, analyze and eliminate outliers, and check for gaps.
3) Carry out data preprocessing (noise removal, normalization, etc.).
4) Train multiple models to predict tensile modulus and tensile strength. When building a model, it is necessary to leave 30% of the data for testing the model, and the rest is where the models are trained. When building models, search for model hyperparameters using grid search with cross-validation, the number of blocks is 10.
5) Write a neural network that will recommend the matrix-filler ratio.
6) Develop an application with a graphical or command line interface that will issue the prediction obtained in task 4 or 5 (one or two predictions, the student's choice).
7) Evaluate the accuracy of the model on the training and test dataset.
8) Create a repository in GitHub / GitLab and place the research code there. Prepare a README file.
Description:
Composite materials are artificially created materials, consisting of several others with a clear boundary between them. Composites have properties that are not observed in the individual components. In this case, composites are a monolithic material, i.e., the components of the material are inseparable from each other without destroying the structure as a whole.At the input there is data on the initial properties of the components of composite materials (the amount of binder, filler, curing temperature, etc.). At the output, it is necessary to predict a number of final properties of the obtained composite materials.
Dataset with properties of composites. Join do by index join type INNER
https://drive.google.com/file/d/1B1s5gBlvgU81H9GGolLQVw_SOi-vyNf2/view?usp=sharing
Required:
1) To study the theoretical foundations and methods for solving the problem.
2) Conduct an exploratory analysis of the proposed data. It is necessary to draw histograms of the distribution of each of the variables, box and whisker plots, pairwise scatter plots. It is also necessary to obtain an average, median value for each column, analyze and eliminate outliers, and check for gaps.
3) Carry out data preprocessing (noise removal, normalization, etc.).
4) Train multiple models to predict tensile modulus and tensile strength. When building a model, it is necessary to leave 30% of the data for testing the model, and the rest is where the models are trained. When building models, search for model hyperparameters using grid search with cross-validation, the number of blocks is 10.
5) Write a neural network that will recommend the matrix-filler ratio.
6) Develop an application with a graphical or command line interface that will issue the prediction obtained in task 4 or 5 (one or two predictions, the student's choice).
7) Evaluate the accuracy of the model on the training and test dataset.
8) Create a repository in GitHub / GitLab and place the research code there. Prepare a README file.
Related categories:
Data Science