The standard way to encode a categorical variable- build one indicator function per value (we don't supress one level). @author johnmount
| 17 | * |
| 18 | */ |
| 19 | public final class LevelIndicators implements VariableMapping { |
| 20 | private static final long serialVersionUID = 1L; |
| 21 | |
| 22 | private final String origColumn; |
| 23 | private final int index; |
| 24 | private final Map<String,Integer> levelCodes = new HashMap<String,Integer>(); |
| 25 | |
| 26 | public LevelIndicators(final String origColumn, final int index, final Collection<String> levels) { |
| 27 | this.origColumn = origColumn; |
| 28 | this.index = index; |
| 29 | for(final String li: new TreeSet<String>(levels)) { |
| 30 | levelCodes.put(li,levelCodes.size()); |
| 31 | } |
| 32 | } |
| 33 | |
| 34 | public Set<String> levels() { |
| 35 | return levelCodes.keySet(); |
| 36 | } |
| 37 | |
| 38 | public int index(final String level) { |
| 39 | return index + levelCodes.get(level); |
| 40 | } |
| 41 | |
| 42 | @Override |
| 43 | public String origColumn() { |
| 44 | return origColumn; |
| 45 | } |
| 46 | |
| 47 | @Override |
| 48 | public int indexL() { |
| 49 | return index; |
| 50 | } |
| 51 | |
| 52 | @Override |
| 53 | public int indexR() { |
| 54 | return index + levelCodes.size(); |
| 55 | } |
| 56 | |
| 57 | @Override |
| 58 | public boolean wantRegularization() { |
| 59 | return false; |
| 60 | } |
| 61 | |
| 62 | @Override |
| 63 | public void process(final BurstMap row, final double[] vec) { |
| 64 | final String level = row.getAsString(origColumn); |
| 65 | if(level!=null) { |
| 66 | final Integer code = levelCodes.get(level); |
| 67 | if(code!=null) { |
| 68 | vec[index+code] = 1.0; |
| 69 | } |
| 70 | } |
| 71 | } |
| 72 | |
| 73 | @Override |
| 74 | public String name() { |
| 75 | return "CategoricalLevels"; |
| 76 | } |
nothing calls this directly
no outgoing calls
no test coverage detected