ML Database

model.conf

// survival        Survival (0 = No; 1 = Yes)
// pclass          Passenger Class (1 = 1st; 2 = 2nd; 3 = 3rd)
// name            Name
// sex             Sex
// age             Age
// sibsp           Number of Siblings/Spouses Aboard
// parch           Number of Parents/Children Aboard
// ticket          Ticket Number
// fare            Passenger Fare
// cabin           Cabin
// embarked        Port of Embarkation


// Binary
// Multiclass
// Regression

// What Type of model are you interested in creating (classification, regression)
classification {
  // which field is our labeled answer set
  target: 'survival',
  // what models should we try
  models: [
    'KNeighborsClassifier', 
    'SVC',
    'DecisionTreeClassifier',
    'RandomForestClassifier',
    'AdaBoostClassifier',
    'GaussianNB'
  ],
  grid_search : {
    'loss': 'rmse'
  }
}

classification {
  target: 'survival',
  models: [
    'LogisticRegression',
    'RandomForest',
    'GaussianNB'
  ],
  // bagging
  // boosting
  // voting
  voting: {
    // majority
    style: 'hard',
    weights [2,1,2]
  }
  gridsearch {
    cross_validations: 5
  }
}

transform.conf

// titanic data set
// survival        Survival (0 = No; 1 = Yes)
// pclass          Passenger Class (1 = 1st; 2 = 2nd; 3 = 3rd)
// name            Name
// sex             Sex
// age             Age
// sibsp           Number of Siblings/Spouses Aboard
// parch           Number of Parents/Children Aboard
// ticket          Ticket Number
// fare            Passenger Fare
// cabin           Cabin
// embarked        Port of Embarkation

// Impute: 
// Imputation. For various reasons, many real world datasets contain missing values, 
// often encoded as blanks, NaNs or other placeholders. Such datasets however are 
// incompatible with scikit-learn estimators which assume that all values in an array 
// are numerical, and that all have and hold meaning.
// Note: This must be done before scaling and other
impute {
  // If a null value is found replace with the mean
  mean: ['age','fare'],
  // If a null value is found replace with a default
  default: ['embarked', 'NYC'] 
}

// Scale:
// Standardization, or mean removal and variance scaling
// Normalization: Normalization is the process of scaling individual samples to have unit norm. 

scale {
  // Convert the following fields into standard deviations from the mean
  zstandard : ['age', 'sibsp', 'parch', 'fare']
  // replace columns with standardized values
  replace: true 
}

// Encode:
// Binarization http://scikit-learn.org/stable/modules/preprocessing.html#binarization
// Encoding Categorical Features
encode {
  // create a sex_male, sex_female, embarked_nyc, embarked_x, embarked_y, ... fields
  onehot: ['sex','embarked'],
  // bin the 
  discretize: {
    fields: ['age'],
    method: 'binning',
    by: 10,
  }
  // Discretization might need to be a concern in the ingestion phase??
}

generate {
  polynomial: ['sibsp','parch']
}