• Create a project.
• Read the data into R as a data frame;
• select one of low-speed sections (speed = 3, or 6);
• select one of normal speed sections (speed = 9, 12, …, 27);
• for each of your selected datasets, you are requested to build machine learning models
to predict Turbine Compressor decay (kMc)
or its categories (kMc.cat) as required as follows.
Part B: Use ML to build models for ship CBM Data set *Data Set will be attached in separate ZIP file* Task Description · Create a project. · Read the data into R as a data frame; · select one of low-speed sections (speed = 3, or 6); · select one of normal speed sections (speed = 9, 12, …, 27); · for each of your selected datasets, you are requested to build machine learning models to predict Turbine Compressor decay (kMc) or its categories (kMc.cat) as required as follows. H1. EDA for a normal speed section (9, …, 27) · In order to build model, you need to conduct EDA to get insight about the relation and consider proper models to use. Show just one or two plots and explain what you found from the plots. H2. Build two models of different type · Split the data into training and test datasets (Do not let training data beyond 70% in this case). · Build two models of different type of your choice, based on your EDA findings. · State what is your response variable (y) and what are predictors (x) (i.e. y ~ x) · Save and submit each of your model as RDS file. · Evaluate each model with proper metrics. · Compare these two models and select one. Explain why. · Note: to get full score for this task, your model needs to have a good performance. L1. EDA for a low-speed section (3 or 6) · In order to build model, you need to conduct EDA to get insight about the relation and consider proper models to use. Show just one or two plots and explain what you found from the plots. L2. Build optimal models for a low-speed section · Split the data into training and test datasets (Do not let training data beyond 70% in this case). · Build two models of different type of your choice, based on your EDA findings. You are required to use caret to optimize at least one of the models. · State what is your response variable (y) and what are predictors (x) (i.e. y ~ x) · State what is the range of your model search, and which is the best. · Save and submit each of your model as RDS file. · Evaluate each model with proper metrics. · Visualize model performance. · Compare these two models and select one. Explain why. • The given data set contains 3000 observations with 16 columns. The following columns in original dataset have been removed, because lever position is simply one to one mapping to speed, each starboard propeller torque is equivalent to port propeller torque, and GT Compressor inlet air temperature and GT Compressor inlet air pressure are just a constant respectively. – lever position (1), – Starboard propeller torque (6), – GT Compressor inlet air temperature (9), – and GT Compressor inlet air pressure (12). • New factor columns “kMc.cat” and “kMt.cat” – Gas Turbine Compressor decay (kMc) and Gas Turbine decay (kMt) are discretized into 6 categories: VVH (very very high decay), VH (very high decay), H (high decay), M (medium level of decay), L(low level of decay), and Non (no decay). •The given data set contains 3000 observations with 16 columns. The following columns in original dataset have been removed, because lever position is simply one to one mapping to speed, each starboard propeller torque is equivalent to port propeller torque, and GT Compressor inlet air temperature and GT Compressor inlet air pressure are just a constant respectively. –lever position (1), –Starboard propeller torque (6), –GT Compressor inlet air temperature (9), –and GT Compressor inlet air pressure (12). •New factor columns “kMc.cat” and “kMt.cat” –Gas Turbine Compressor decay (kMc) and Gas Turbine decay (kMt) are discretized into 6 categories: VVH (very very high decay), VH (very high decay), H (high decay), M (medium level of decay), L(low level of decay), and Non (no decay).