[13] | 1 | package de.ugoe.cs.eventbench.models;
|
---|
[12] | 2 |
|
---|
[93] | 3 | import java.security.InvalidParameterException;
|
---|
| 4 | import java.util.ArrayList;
|
---|
[102] | 5 | import java.util.Collection;
|
---|
[252] | 6 | import java.util.HashSet;
|
---|
[93] | 7 | import java.util.LinkedHashSet;
|
---|
[12] | 8 | import java.util.LinkedList;
|
---|
| 9 | import java.util.List;
|
---|
| 10 | import java.util.Random;
|
---|
[80] | 11 | import java.util.Set;
|
---|
[12] | 12 |
|
---|
| 13 | import de.ugoe.cs.eventbench.data.Event;
|
---|
[23] | 14 | import de.ugoe.cs.eventbench.models.Trie.Edge;
|
---|
| 15 | import de.ugoe.cs.eventbench.models.Trie.TrieVertex;
|
---|
| 16 | import edu.uci.ics.jung.graph.Tree;
|
---|
[12] | 17 |
|
---|
[100] | 18 | /**
|
---|
| 19 | * <p>
|
---|
| 20 | * Implements a skeleton for stochastic processes that can calculate
|
---|
| 21 | * probabilities based on a trie. The skeleton provides all functionalities of
|
---|
| 22 | * {@link IStochasticProcess} except
|
---|
| 23 | * {@link IStochasticProcess#getProbability(List, Event)}.
|
---|
| 24 | * </p>
|
---|
| 25 | *
|
---|
| 26 | * @author Steffen Herbold
|
---|
| 27 | * @version 1.0
|
---|
| 28 | */
|
---|
[17] | 29 | public abstract class TrieBasedModel implements IStochasticProcess {
|
---|
[12] | 30 |
|
---|
[86] | 31 | /**
|
---|
[100] | 32 | * <p>
|
---|
[86] | 33 | * Id for object serialization.
|
---|
[100] | 34 | * </p>
|
---|
[86] | 35 | */
|
---|
| 36 | private static final long serialVersionUID = 1L;
|
---|
| 37 |
|
---|
[100] | 38 | /**
|
---|
| 39 | * <p>
|
---|
| 40 | * The order of the trie, i.e., the maximum length of subsequences stored in
|
---|
| 41 | * the trie.
|
---|
| 42 | * </p>
|
---|
| 43 | */
|
---|
[16] | 44 | protected int trieOrder;
|
---|
[12] | 45 |
|
---|
[100] | 46 | /**
|
---|
| 47 | * <p>
|
---|
| 48 | * Trie on which the probability calculations are based.
|
---|
| 49 | * </p>
|
---|
| 50 | */
|
---|
[182] | 51 | protected Trie<Event<?>> trie = null;
|
---|
[100] | 52 |
|
---|
| 53 | /**
|
---|
| 54 | * <p>
|
---|
| 55 | * Random number generator used by probabilistic sequence generation
|
---|
| 56 | * methods.
|
---|
| 57 | * </p>
|
---|
| 58 | */
|
---|
[12] | 59 | protected final Random r;
|
---|
| 60 |
|
---|
[100] | 61 | /**
|
---|
| 62 | * <p>
|
---|
| 63 | * Constructor. Creates a new TrieBasedModel that can be used for stochastic
|
---|
| 64 | * processes with a Markov order less than or equal to {@code markovOrder}.
|
---|
| 65 | * </p>
|
---|
| 66 | *
|
---|
| 67 | * @param markovOrder
|
---|
| 68 | * Markov order of the model
|
---|
| 69 | * @param r
|
---|
| 70 | * random number generator used by probabilistic methods of the
|
---|
| 71 | * class
|
---|
[342] | 72 | * @throws InvalidParameterException
|
---|
| 73 | * thrown if markovOrder is less than 0 or the random number
|
---|
| 74 | * generator r is null
|
---|
[100] | 75 | */
|
---|
[16] | 76 | public TrieBasedModel(int markovOrder, Random r) {
|
---|
[12] | 77 | super();
|
---|
[342] | 78 | if (markovOrder < 0) {
|
---|
| 79 | throw new InvalidParameterException(
|
---|
| 80 | "markov order must not be less than 0");
|
---|
| 81 | }
|
---|
| 82 | if (r == null) {
|
---|
| 83 | throw new InvalidParameterException(
|
---|
| 84 | "random number generator r must not be null");
|
---|
| 85 | }
|
---|
[100] | 86 | this.trieOrder = markovOrder + 1;
|
---|
[12] | 87 | this.r = r;
|
---|
| 88 | }
|
---|
| 89 |
|
---|
[100] | 90 | /**
|
---|
| 91 | * <p>
|
---|
| 92 | * Trains the model by generating a trie from which probabilities are
|
---|
[182] | 93 | * calculated. The trie is newly generated based solely on the passed
|
---|
| 94 | * sequences. If an existing model should only be updated, use
|
---|
| 95 | * {@link #update(Collection)} instead.
|
---|
[100] | 96 | * </p>
|
---|
| 97 | *
|
---|
| 98 | * @param sequences
|
---|
| 99 | * training data
|
---|
[342] | 100 | * @throws InvalidParameterException
|
---|
| 101 | * thrown is sequences is null
|
---|
[100] | 102 | */
|
---|
[325] | 103 | public void train(Collection<List<? extends Event<?>>> sequences) {
|
---|
[182] | 104 | trie = null;
|
---|
| 105 | update(sequences);
|
---|
| 106 | }
|
---|
[100] | 107 |
|
---|
[182] | 108 | /**
|
---|
| 109 | * <p>
|
---|
| 110 | * Trains the model by updating the trie from which the probabilities are
|
---|
| 111 | * calculated. This function updates an existing trie. In case no trie
|
---|
| 112 | * exists yet, a new trie is generated and the function behaves like
|
---|
| 113 | * {@link #train(Collection)}.
|
---|
| 114 | * </p>
|
---|
| 115 | *
|
---|
| 116 | * @param sequences
|
---|
| 117 | * training data
|
---|
[342] | 118 | * @throws InvalidParameterException
|
---|
| 119 | * thrown is sequences is null
|
---|
[182] | 120 | */
|
---|
[325] | 121 | public void update(Collection<List<? extends Event<?>>> sequences) {
|
---|
[252] | 122 | if (sequences == null) {
|
---|
[342] | 123 | throw new InvalidParameterException("sequences must not be null");
|
---|
[252] | 124 | }
|
---|
[182] | 125 | if (trie == null) {
|
---|
| 126 | trie = new Trie<Event<?>>();
|
---|
| 127 | }
|
---|
[325] | 128 | for (List<? extends Event<?>> sequence : sequences) {
|
---|
[100] | 129 | List<Event<?>> currentSequence = new LinkedList<Event<?>>(sequence); // defensive
|
---|
| 130 | // copy
|
---|
[12] | 131 | currentSequence.add(0, Event.STARTEVENT);
|
---|
| 132 | currentSequence.add(Event.ENDEVENT);
|
---|
[100] | 133 |
|
---|
[16] | 134 | trie.train(currentSequence, trieOrder);
|
---|
[12] | 135 | }
|
---|
| 136 | }
|
---|
| 137 |
|
---|
[100] | 138 | /*
|
---|
| 139 | * (non-Javadoc)
|
---|
| 140 | *
|
---|
[17] | 141 | * @see de.ugoe.cs.eventbench.models.IStochasticProcess#randomSequence()
|
---|
| 142 | */
|
---|
| 143 | @Override
|
---|
[12] | 144 | public List<? extends Event<?>> randomSequence() {
|
---|
| 145 | List<Event<?>> sequence = new LinkedList<Event<?>>();
|
---|
[342] | 146 | if (trie != null) {
|
---|
[252] | 147 | IncompleteMemory<Event<?>> context = new IncompleteMemory<Event<?>>(
|
---|
| 148 | trieOrder - 1);
|
---|
| 149 | context.add(Event.STARTEVENT);
|
---|
[342] | 150 |
|
---|
[252] | 151 | Event<?> currentState = Event.STARTEVENT;
|
---|
[342] | 152 |
|
---|
[252] | 153 | boolean endFound = false;
|
---|
[342] | 154 |
|
---|
[252] | 155 | while (!endFound) {
|
---|
| 156 | double randVal = r.nextDouble();
|
---|
| 157 | double probSum = 0.0;
|
---|
| 158 | List<Event<?>> currentContext = context.getLast(trieOrder);
|
---|
| 159 | for (Event<?> symbol : trie.getKnownSymbols()) {
|
---|
| 160 | probSum += getProbability(currentContext, symbol);
|
---|
| 161 | if (probSum >= randVal) {
|
---|
| 162 | endFound = (symbol == Event.ENDEVENT);
|
---|
| 163 | if (!(symbol == Event.STARTEVENT || symbol == Event.ENDEVENT)) {
|
---|
[342] | 164 | // only add the symbol the sequence if it is not
|
---|
| 165 | // START
|
---|
[252] | 166 | // or END
|
---|
| 167 | context.add(symbol);
|
---|
| 168 | currentState = symbol;
|
---|
| 169 | sequence.add(currentState);
|
---|
| 170 | }
|
---|
| 171 | break;
|
---|
[12] | 172 | }
|
---|
| 173 | }
|
---|
| 174 | }
|
---|
| 175 | }
|
---|
| 176 | return sequence;
|
---|
| 177 | }
|
---|
[100] | 178 |
|
---|
| 179 | /**
|
---|
| 180 | * <p>
|
---|
| 181 | * Returns a Dot representation of the internal trie.
|
---|
| 182 | * </p>
|
---|
| 183 | *
|
---|
| 184 | * @return dot representation of the internal trie
|
---|
| 185 | */
|
---|
[30] | 186 | public String getTrieDotRepresentation() {
|
---|
[252] | 187 | if (trie == null) {
|
---|
| 188 | return "";
|
---|
| 189 | } else {
|
---|
| 190 | return trie.getDotRepresentation();
|
---|
| 191 | }
|
---|
[30] | 192 | }
|
---|
[100] | 193 |
|
---|
| 194 | /**
|
---|
| 195 | * <p>
|
---|
| 196 | * Returns a {@link Tree} of the internal trie that can be used for
|
---|
| 197 | * visualization.
|
---|
| 198 | * </p>
|
---|
| 199 | *
|
---|
| 200 | * @return {@link Tree} depicting the internal trie
|
---|
| 201 | */
|
---|
[23] | 202 | public Tree<TrieVertex, Edge> getTrieGraph() {
|
---|
[252] | 203 | if (trie == null) {
|
---|
| 204 | return null;
|
---|
| 205 | } else {
|
---|
| 206 | return trie.getGraph();
|
---|
| 207 | }
|
---|
[23] | 208 | }
|
---|
[12] | 209 |
|
---|
[100] | 210 | /**
|
---|
| 211 | * <p>
|
---|
| 212 | * The string representation of the model is {@link Trie#toString()} of
|
---|
| 213 | * {@link #trie}.
|
---|
| 214 | * </p>
|
---|
| 215 | *
|
---|
| 216 | * @see java.lang.Object#toString()
|
---|
| 217 | */
|
---|
[12] | 218 | @Override
|
---|
| 219 | public String toString() {
|
---|
[252] | 220 | if (trie == null) {
|
---|
| 221 | return "";
|
---|
| 222 | } else {
|
---|
| 223 | return trie.toString();
|
---|
| 224 | }
|
---|
[12] | 225 | }
|
---|
[100] | 226 |
|
---|
| 227 | /*
|
---|
| 228 | * (non-Javadoc)
|
---|
| 229 | *
|
---|
| 230 | * @see de.ugoe.cs.eventbench.models.IStochasticProcess#getNumStates()
|
---|
| 231 | */
|
---|
| 232 | @Override
|
---|
[129] | 233 | public int getNumSymbols() {
|
---|
[252] | 234 | if (trie == null) {
|
---|
| 235 | return 0;
|
---|
| 236 | } else {
|
---|
| 237 | return trie.getNumSymbols();
|
---|
| 238 | }
|
---|
[66] | 239 | }
|
---|
[100] | 240 |
|
---|
| 241 | /*
|
---|
| 242 | * (non-Javadoc)
|
---|
| 243 | *
|
---|
| 244 | * @see de.ugoe.cs.eventbench.models.IStochasticProcess#getStateStrings()
|
---|
| 245 | */
|
---|
| 246 | @Override
|
---|
[129] | 247 | public String[] getSymbolStrings() {
|
---|
[252] | 248 | if (trie == null) {
|
---|
| 249 | return new String[0];
|
---|
| 250 | }
|
---|
[129] | 251 | String[] stateStrings = new String[getNumSymbols()];
|
---|
[100] | 252 | int i = 0;
|
---|
| 253 | for (Event<?> symbol : trie.getKnownSymbols()) {
|
---|
[70] | 254 | stateStrings[i] = symbol.toString();
|
---|
| 255 | i++;
|
---|
| 256 | }
|
---|
| 257 | return stateStrings;
|
---|
| 258 | }
|
---|
[100] | 259 |
|
---|
| 260 | /*
|
---|
| 261 | * (non-Javadoc)
|
---|
| 262 | *
|
---|
| 263 | * @see de.ugoe.cs.eventbench.models.IStochasticProcess#getEvents()
|
---|
| 264 | */
|
---|
| 265 | @Override
|
---|
[102] | 266 | public Collection<? extends Event<?>> getEvents() {
|
---|
[252] | 267 | if (trie == null) {
|
---|
| 268 | return new HashSet<Event<?>>();
|
---|
| 269 | } else {
|
---|
| 270 | return trie.getKnownSymbols();
|
---|
| 271 | }
|
---|
[80] | 272 | }
|
---|
[100] | 273 |
|
---|
| 274 | /*
|
---|
| 275 | * (non-Javadoc)
|
---|
| 276 | *
|
---|
| 277 | * @see
|
---|
| 278 | * de.ugoe.cs.eventbench.models.IStochasticProcess#generateSequences(int)
|
---|
| 279 | */
|
---|
| 280 | @Override
|
---|
[102] | 281 | public Collection<List<? extends Event<?>>> generateSequences(int length) {
|
---|
[94] | 282 | return generateSequences(length, false);
|
---|
[93] | 283 | }
|
---|
[100] | 284 |
|
---|
| 285 | /*
|
---|
| 286 | * (non-Javadoc)
|
---|
| 287 | *
|
---|
| 288 | * @see
|
---|
| 289 | * de.ugoe.cs.eventbench.models.IStochasticProcess#generateSequences(int,
|
---|
| 290 | * boolean)
|
---|
| 291 | */
|
---|
| 292 | @Override
|
---|
| 293 | public Set<List<? extends Event<?>>> generateSequences(int length,
|
---|
| 294 | boolean fromStart) {
|
---|
| 295 | Set<List<? extends Event<?>>> sequenceSet = new LinkedHashSet<List<? extends Event<?>>>();
|
---|
| 296 | if (length < 1) {
|
---|
| 297 | throw new InvalidParameterException(
|
---|
| 298 | "Length of generated subsequences must be at least 1.");
|
---|
[94] | 299 | }
|
---|
[100] | 300 | if (length == 1) {
|
---|
| 301 | if (fromStart) {
|
---|
[94] | 302 | List<Event<?>> subSeq = new LinkedList<Event<?>>();
|
---|
| 303 | subSeq.add(Event.STARTEVENT);
|
---|
[95] | 304 | sequenceSet.add(subSeq);
|
---|
[94] | 305 | } else {
|
---|
[100] | 306 | for (Event<?> event : getEvents()) {
|
---|
[94] | 307 | List<Event<?>> subSeq = new LinkedList<Event<?>>();
|
---|
| 308 | subSeq.add(event);
|
---|
| 309 | sequenceSet.add(subSeq);
|
---|
| 310 | }
|
---|
| 311 | }
|
---|
| 312 | return sequenceSet;
|
---|
| 313 | }
|
---|
[102] | 314 | Collection<? extends Event<?>> events = getEvents();
|
---|
| 315 | Collection<List<? extends Event<?>>> seqsShorter = generateSequences(
|
---|
[100] | 316 | length - 1, fromStart);
|
---|
| 317 | for (Event<?> event : events) {
|
---|
| 318 | for (List<? extends Event<?>> seqShorter : seqsShorter) {
|
---|
[94] | 319 | Event<?> lastEvent = event;
|
---|
[100] | 320 | if (getProbability(seqShorter, lastEvent) > 0.0) {
|
---|
[94] | 321 | List<Event<?>> subSeq = new ArrayList<Event<?>>(seqShorter);
|
---|
| 322 | subSeq.add(lastEvent);
|
---|
| 323 | sequenceSet.add(subSeq);
|
---|
| 324 | }
|
---|
| 325 | }
|
---|
| 326 | }
|
---|
| 327 | return sequenceSet;
|
---|
| 328 | }
|
---|
[100] | 329 |
|
---|
| 330 | /*
|
---|
| 331 | * (non-Javadoc)
|
---|
| 332 | *
|
---|
| 333 | * @see
|
---|
| 334 | * de.ugoe.cs.eventbench.models.IStochasticProcess#generateValidSequences
|
---|
| 335 | * (int)
|
---|
| 336 | */
|
---|
| 337 | @Override
|
---|
[118] | 338 | public Collection<List<? extends Event<?>>> generateValidSequences(
|
---|
| 339 | int length) {
|
---|
[94] | 340 | // check for min-length implicitly done by generateSequences
|
---|
[118] | 341 | Collection<List<? extends Event<?>>> allSequences = generateSequences(
|
---|
| 342 | length, true);
|
---|
[102] | 343 | Collection<List<? extends Event<?>>> validSequences = new LinkedHashSet<List<? extends Event<?>>>();
|
---|
[100] | 344 | for (List<? extends Event<?>> sequence : allSequences) {
|
---|
| 345 | if (sequence.size() == length
|
---|
| 346 | && Event.ENDEVENT.equals(sequence.get(sequence.size() - 1))) {
|
---|
[95] | 347 | validSequences.add(sequence);
|
---|
[94] | 348 | }
|
---|
| 349 | }
|
---|
| 350 | return validSequences;
|
---|
| 351 | }
|
---|
[12] | 352 |
|
---|
[118] | 353 | /*
|
---|
| 354 | * (non-Javadoc)
|
---|
| 355 | *
|
---|
| 356 | * @see
|
---|
| 357 | * de.ugoe.cs.eventbench.models.IStochasticProcess#getProbability(java.util
|
---|
| 358 | * .List)
|
---|
| 359 | */
|
---|
| 360 | @Override
|
---|
| 361 | public double getProbability(List<? extends Event<?>> sequence) {
|
---|
[342] | 362 | if (sequence == null) {
|
---|
| 363 | throw new InvalidParameterException("sequence must not be null");
|
---|
| 364 | }
|
---|
[118] | 365 | double prob = 1.0;
|
---|
[342] | 366 | List<Event<?>> context = new LinkedList<Event<?>>();
|
---|
| 367 | for (Event<?> event : sequence) {
|
---|
| 368 | prob *= getProbability(context, event);
|
---|
| 369 | context.add(event);
|
---|
[118] | 370 | }
|
---|
| 371 | return prob;
|
---|
| 372 | }
|
---|
| 373 |
|
---|
[182] | 374 | /*
|
---|
| 375 | * (non-Javadoc)
|
---|
| 376 | *
|
---|
[129] | 377 | * @see de.ugoe.cs.eventbench.models.IStochasticProcess#getNumFOMStates()
|
---|
| 378 | */
|
---|
| 379 | @Override
|
---|
| 380 | public int getNumFOMStates() {
|
---|
[252] | 381 | if (trie == null) {
|
---|
| 382 | return 0;
|
---|
| 383 | } else {
|
---|
| 384 | return trie.getNumLeafAncestors();
|
---|
| 385 | }
|
---|
[129] | 386 | }
|
---|
[248] | 387 |
|
---|
| 388 | /*
|
---|
| 389 | * (non-Javadoc)
|
---|
| 390 | *
|
---|
| 391 | * @see de.ugoe.cs.eventbench.models.IStochasticProcess#getNumTransitions()
|
---|
| 392 | */
|
---|
| 393 | @Override
|
---|
| 394 | public int getNumTransitions() {
|
---|
[252] | 395 | if (trie == null) {
|
---|
| 396 | return 0;
|
---|
| 397 | } else {
|
---|
| 398 | return trie.getNumLeafs();
|
---|
| 399 | }
|
---|
[248] | 400 | }
|
---|
[12] | 401 | } |
---|