Submit your solution codes into a notebook file with “.ipynb” extension. Write discussions and explanations including outputs and figures into a separate file and submit as a PDF file. 1. What are the...

Have attached data set , data description and instructions


Submit your solution codes into a notebook file with “.ipynb” extension. Write discussions and explanations including outputs and figures into a separate file and submit as a PDF file. 1. What are the differences between hyperparameter and parameter of a machine learning (ML) model. Explain your answer using at least two machine learning models that you have learned in this unit. 2. Prove that Elastic net can be used as either LASSO or Ridge regulariser. Background The recently started human and other genome projects are likely to change the situation of molecular biology. Comprehensive analyses of whole genomic sequences will enable us to understand the general mechanisms of how protein and nucleic acid functions are encoded in the sequence data. Dataset filename: yeast2vs4.csv Dataset description: There are 8 features and one target in the dataset. All the features are in a numerical format, and the target is in text format. For further information about the attributes, please read “Data Set Information.pdf”. 3. Analyse the importance of the features for predicting presence or absence of protein using two different approaches. Explain the similarity/difference between outcomes. 4. Create three supervised machine learning (ML) models except any ensemble approach for predicting presence or absence of protein. a. Report performance score using a suitable metric. Is it possible that the presented result is an overfitted one? Justify. b. Justify different design decisions for each ML model used to answer this question. c. Have you optimised any hyper-parameters for each ML model? What are they? Why have you done that? Explain. d. Finally, make a recommendation based on the reported results and justify it. 5. Build three ensemble models for predicting presence or absence of protein. a. When do you want to use ensemble models over other ML models? b. What are the similarities or differences between these models? c. Is there any preferable scenario for using any specific model among the set of ensemble models? d. Write a report comparing performances of models built in question 5 and 6. Report the best method based on model complexity and performance. e. Is it possible to build ensemble model using ML classifiers other than decision tree? If yes, then explain with an example. Microsoft Word - Data Set Information.docx Attribute Information: 1. mcg: McGeoch's method for signal sequence recognition. 2. gvh: von Heijne's method for signal sequence recognition. 3. alm: Score of the ALOM membrane spanning region prediction program. 4. mit: Score of discriminant analysis of the amino acid content of the N-terminal region (20 residues long) of mitochondrial and non-mitochondrial proteins. 5. erl: Presence of "HDEL" substring (thought to act as a signal for retention in the endoplasmic reticulum lumen). Binary attribute. 6. pox: Peroxisomal targeting signal in the C-terminus. 7. vac: Score of discriminant analysis of the amino acid content of vacuolar and extracellular proteins. 8. nuc: Score of discriminant analysis of nuclear localization signals of nuclear and non-nuclear proteins. 9. class: Presence or absence of protein {positive, negative}. Further information: Following are the articles that have used this dataset: 1. "Expert Sytem for Predicting Protein Localization Sites in Gram-Negative Bacteria", Kenta Nakai & Minoru Kanehisa, PROTEINS: Structure, Function, and Genetics 11:95-110, 1991. 2. "A Knowledge Base for Predicting Protein Localization Sites in Eukaryotic Cells", Kenta Nakai & Minoru Kanehisa, Genomics 14:897-911, 1992.