Examples with MapReducerStatsWorker ml.shifu.shifu.core.processor.stats.MapReducerStatsWorker used on opensource projects

Example 1 with MapReducerStatsWorker

use of ml.shifu.shifu.core.processor.stats.MapReducerStatsWorker in project shifu by ShifuML.

the class StatsModelProcessor method run.

/**
 * runner for statistics
 */
@Override
public int run() throws Exception {
    log.info("Step Start: stats");
    long start = System.currentTimeMillis();
    try {
        // 0. set up and sync to HDFS
        setUp(ModelStep.STATS);
        // resync ModelConfig.json/ColumnConfig.json to HDFS
        syncDataToHdfs(modelConfig.getDataSet().getSource());
        if (getBooleanParam(this.params, Constants.IS_COMPUTE_CORR)) {
            // 1. validate if run stats before run stats -correlation
            boolean foundValidMeanValueColumn = isMeanCalculated();
            if (!foundValidMeanValueColumn) {
                log.warn("Some mean value of column is null, could you check if you run 'shifu stats'.");
                return -1;
            }
            // 2. compute correlation
            log.info("Start computing correlation value ...");
            SourceType source = this.modelConfig.getDataSet().getSource();
            String corrPath = super.getPathFinder().getCorrelationPath(source);
            // check if can start from existing output
            boolean reuseCorrResult = Environment.getBoolean("shifu.stats.corr.reuse", Boolean.FALSE);
            if (reuseCorrResult && ShifuFileUtils.isFileExists(corrPath, SourceType.HDFS)) {
                dumpAndCalculateCorrelationResult(source, corrPath);
            } else {
                runCorrMapReduceJob();
            }
            // 3. save column config list
            saveColumnConfigList();
        } else if (getBooleanParam(this.params, Constants.IS_COMPUTE_PSI)) {
            boolean foundValidMeanValueColumn = isMeanCalculated();
            if (!foundValidMeanValueColumn) {
                log.warn("Some mean value of column is null, could you check if you run 'shifu stats'.");
                return -1;
            }
            if (StringUtils.isNotEmpty(modelConfig.getPsiColumnName())) {
                new MapReducerStatsWorker(this, modelConfig, columnConfigList).runPSI();
                // save column config list after running PSI successfully
                saveColumnConfigList();
            } else {
                log.warn("To Run PSI please set your PSI column in dataSet::psiColumnName.");
            }
        } else if (getBooleanParam(this.params, Constants.IS_REBIN)) {
            // run the re-binning
            String backupColumnConfigPath = this.pathFinder.getBackupColumnConfig();
            if (!ShifuFileUtils.isFileExists(new Path(backupColumnConfigPath), SourceType.LOCAL)) {
                ShifuFileUtils.createDirIfNotExists(new SourceFile(Constants.TMP, SourceType.LOCAL));
                saveColumnConfigList(backupColumnConfigPath, this.columnConfigList);
            } else {
                // existing backup ColumnConfig.json, use binning info in it to do rebin
                List<ColumnConfig> backColumnConfigList = CommonUtils.loadColumnConfigList(backupColumnConfigPath, SourceType.LOCAL, false);
                for (ColumnConfig backupColumnConfig : backColumnConfigList) {
                    for (ColumnConfig columnConfig : this.columnConfigList) {
                        if (NSColumnUtils.isColumnEqual(backupColumnConfig.getColumnName(), columnConfig.getColumnName())) {
                            columnConfig.setColumnBinning(backupColumnConfig.getColumnBinning());
                        }
                    }
                }
            }
            // user provide candidate variable list or not
            boolean hasCandidates = CommonUtils.hasCandidateColumns(this.columnConfigList);
            List<ColumnConfig> rebinColumns = new ArrayList<ColumnConfig>();
            List<String> catVariables = getStringList(this.params, Constants.REQUEST_VARS, ",");
            for (ColumnConfig columnConfig : this.columnConfigList) {
                if (CollectionUtils.isEmpty(catVariables) || isRequestColumn(catVariables, columnConfig)) {
                    if (CommonUtils.isGoodCandidate(columnConfig, hasCandidates)) {
                        rebinColumns.add(columnConfig);
                    } else {
                        log.warn("Column - {} is not a good candidate. Skip it.", columnConfig.getColumnName());
                    }
                }
            }
            if (CollectionUtils.isNotEmpty(rebinColumns)) {
                for (ColumnConfig columnConfig : rebinColumns) {
                    doReBin(columnConfig);
                }
            }
            // use the merge ColumnConfig.json to replace current one
            saveColumnConfigList();
        } else {
            AbstractStatsExecutor statsExecutor = null;
            if (modelConfig.isMapReduceRunMode()) {
                if (modelConfig.getBinningAlgorithm().equals(ModelStatsConf.BinningAlgorithm.DynamicBinning)) {
                    statsExecutor = new DIBStatsExecutor(this, modelConfig, columnConfigList);
                } else if (modelConfig.getBinningAlgorithm().equals(ModelStatsConf.BinningAlgorithm.MunroPat)) {
                    statsExecutor = new MunroPatStatsExecutor(this, modelConfig, columnConfigList);
                } else if (modelConfig.getBinningAlgorithm().equals(ModelStatsConf.BinningAlgorithm.MunroPatI)) {
                    statsExecutor = new MunroPatIStatsExecutor(this, modelConfig, columnConfigList);
                } else if (modelConfig.getBinningAlgorithm().equals(ModelStatsConf.BinningAlgorithm.SPDT)) {
                    statsExecutor = new SPDTStatsExecutor(this, modelConfig, columnConfigList);
                } else if (modelConfig.getBinningAlgorithm().equals(ModelStatsConf.BinningAlgorithm.SPDTI)) {
                    statsExecutor = new SPDTIStatsExecutor(this, modelConfig, columnConfigList);
                } else {
                    statsExecutor = new SPDTIStatsExecutor(this, modelConfig, columnConfigList);
                }
            } else if (modelConfig.isLocalRunMode()) {
                statsExecutor = new AkkaStatsWorker(this, modelConfig, columnConfigList);
            } else {
                throw new ShifuException(ShifuErrorCode.ERROR_UNSUPPORT_MODE);
            }
            statsExecutor.doStats();
            // update the backup ColumnConfig.json after running stats
            String backupColumnConfigPath = this.pathFinder.getBackupColumnConfig();
            ShifuFileUtils.createDirIfNotExists(new SourceFile(Constants.TMP, SourceType.LOCAL));
            saveColumnConfigList(backupColumnConfigPath, this.columnConfigList);
        }
        // back up current column config each time as stats will always change CC.json
        this.backupCurrentColumnConfigToLocal(SDF.format(new Date()));
        syncDataToHdfs(modelConfig.getDataSet().getSource());
        clearUp(ModelStep.STATS);
    } catch (ShifuException e) {
        log.error("Error:" + e.getError().toString() + "; msg:" + e.getMessage(), e);
        return -1;
    } catch (Exception e) {
        log.error("Error:" + e.getMessage(), e);
        return -1;
    }
    log.info("Step Finished: stats with {} ms", (System.currentTimeMillis() - start));
    return 0;
}