What’s New in hana.ml.r v2.17.240111

Bug Fixed:

  • Resolved the issue preventing the use of a generic function ‘CallPalAutoWithConnection’ to call native PAL SQL Procedure.
  • Rectified the ‘evaluation.metric’ check failure that occurred when ‘family’ was set to ordinal in ‘hanaml.GLM’.
  • Corrected the precompute parameters check in massive mode for ‘hanaml.UnifiedClustering’.

What’s New in hana.ml.r v2.17.231018

Enhancement:

  • Enhanced the support of data type of DECIMAL.

Bug Fixed:

  • Fixed the type check issue of pmml.export in hanaml.LogisticRegression.

What’s New in hana.ml.r v2.17.230615

New Functions:

  • Added a new function hanaml.ROCKET.
  • Added a new function hanaml.ClassificationTS for time series Classification.

Enhancement:

  • Enhanced precalculated/precomputed distances Matrix input for KMedoid in hanaml.UnifiedClustering.
  • Enabled interval prediction for Linear Regression, GLM and Random Decision Trees in UnifiedRegression.
  • Added SPEC as a new accuracy measure for time-series forecasting.
  • Enhanced auto-ml with reason code and successive halving.

What’s New in hana.ml.r v2.16.230316

New Functions:

  • Added a new function hanaml.Periodogram.
  • Added a Long-Term Series Forecasting method hanaml.LTSF.
  • Added a new function hanaml.AutomaticTimeSeries.

Enhancement:

  • Enhanced hanaml.OutlierDetectionTS with new outlier.method “isolationforest”, “dbscan”.
  • Enhanced hanaml.UnifiedClustering with Massive mode.
  • Enhanced SVM and MLP in Unified Regression with sha/hyperband params.
  • Enhanced automl with new parameter model.table.name to define the table name
  • Enabled sha/hyperband param select for NB, MLP, FRM and ALS.
  • Enhanced HGBT in Unified Classification with early stop parameter.

Bug Fixed:

  • Correction for warn suppression in DataFrame Collect.

What’s New in hana.ml.r v2.15.221216

New Functions:

  • Added time series outlier detection called hanaml.OutlierDetectionTS.
  • Added Kolmogorov-Smirnov Test called hanaml.KsTest.
  • Added time series imputation called hanaml.ImputeTS.

Enhancement:

  • Enhanced text mining with French and Russian support.
  • Enhanced JSON model format support in hanaml.LogisticRegression and hanaml.LinearRegression.
  • Ehhanced hanaml.Correlation with confidence interval calculation.
  • Enhanced the support of pre-defined period setting in hanaml.SeasonalDecompose with a new parameter ‘periods’.
  • Enhanced HGBT, KNN, SVM and MLR with Successive halving.
  • Enhanced auto-ml with multiple config_dict templates.
  • Enhanced multi-class LogisticRegression with AFL state.

API Changes:

  • Added a parameter “json.export” in hanaml.LogisticRegression and hanaml.LinearRegression for exporting model in JSON format.
  • Added new parameters “calculate.confint”, “alpha”, “bartlett” in hanaml.Correlation for confidence interval calculation.
  • Added a parameter ‘periods’ in hanaml.SeasonalDecompose.

What’s New in hana.ml.r v2.14.220918

New Functions:

  • Added Stationarity Test called hanaml.StationarityTest.
  • Added AutoML called hanaml.AutomaticClassification and hanaml.AutomaticRegression.
  • Added Croston TSB method.

Enhancement:

  • Enhanced missing value handling in hanaml.UnifiedClassification and hanaml.UnifiedRegression.
  • Enhanced hanaml.UnifiedClassification and hanaml.UnifiedRegression with Massive mode.
  • Enhanced additive model forecst, ARIMA, AutoARIMA with massive mode.
  • Enhanced garch to support more model types.
  • Enhanced text mining with Spanish language.

Bug Fixed:

  • Fixed null data check in isolation forest.
  • Fixed parameter mapping error in UnifiedClassification.
  • Fixed the bug for parsing ‘c’ param in svm for unified interfaces.
  • Fixed the incompatibleness between afl state enabling and ModelStorage.

What’s New in hana.ml.r v2.13.220511

New Functions:

  • Added Isolation Forest called hanaml.IsolationForest.
  • Added AFL state in unified API.

Enhancement:

  • Enhanced hanaml.WhiteNoiseTest with a new parameter “model.df” to specify the degrees of freedom occuppied by a model.
  • Enhanced HGBT with more loss functions.
  • Enhanced discrete wavelet transform with threshold method and compression.
  • Enhanced text mining with German support.
  • Enhanced expalinability in mechanism LSTM/ATTENTION.

API Changes:

  • Added a new parameter “model.df” in hanaml.WhiteNoiseTest for selecting the degree of freedom.

What’s New in hana.ml.r v2.12.220325

New Functions:

  • Added PCA categorical support.
  • Added Bayesian Structured Time Series(BSTS) forecast.
  • Added Feature Selection.
  • Added Continous HybridGradientBoostingTrees.
  • Added pivot in dataframe.

Enhancement:

  • Supports Explainer in hanaml.ARIMA and hanaml.AutoARIMA.
  • Supports Explainer in hanaml.AdditiveModelForecast.
  • Supports warm start mode in hanaml.HGBTClassifier and hanaml.HGBTRegressor.
  • Supports data compression for discrete wavelet transform and wavelet packet transform.
  • Supports feature-wise explain mode in hanaml.GRUAttention.
  • Supports more types of objective functions in hanaml.HGBT.
  • Supports reference column and ordering as inputs when adding an ID column for DataFrame.
  • Supports precomputed affinity in hanaml.AgglomerateHierarchical.
  • Supports window in hanaml.FFT.

API Changes:

  • Added a new parameter “background.size” in hanaml.ARIMA and AutoARIMA for generating explainer in predict.ARIMA and predict.AutoARIMA.
  • Added new parameters “show.explainer”, “thread.ratio”, “top.k.attributions”, “trend.mod”, “trend.width”, “seasonal.width” in predict.ARIMA and predict.AutoARIMA.
  • Added new parameters “show.explainer”, “decompose.seasonality”, “decompose.holiday” in predict.AdditiveModelForecast.
  • Added new parameters “model” and “warm.start” in hanaml.HGBTClassifier and hanaml.HGBTRegressor to support warm start mode.
  • Added new parameters “compression”, “method”, “threshold” and “level.thresholds”(only valid for hanaml.DWT) in hanaml.DWT and hanaml.WPT.
  • Added new parameter “explain.mode” in hanaml.GRUAttention.
  • Added new parameters “ref.col” and “order” for the AddID() function in DataFrame.
  • Added new parameters “window”, “window.start”, “window.length”, “alpha”, “beta”, “attenuation”, “flattop.mode”, “flattop.precision” and “tukey.r” in hanaml.FFT.

Bug Fixed:

  • Fixed the bug of inconsistent mappings of solvers between logistic regression and multi-class logistic regression.

What’s New in hana.ml.r v2.11.211211

New Functions:

  • Added unifed exponential smoothing.
  • Added GARCH.
  • Added spectral clustering.
  • Added LSTM and attention.
  • Added generic DTW.
  • Added wavelet transforms.

Enhancement:

  • Supports more distributions in MCMC.

API Changes:

  • Added categorical support in SMOTE/SMOTETomek.

What’s New in hana.ml.r v2.10.210918

New Functions:

  • TextMining : TF.Analysis, Text.Classification, Get.Related.Doc, Get.Relevant.Doc, Get.Related.Term, Get.Relevant.Term, Get.Suggested.Term, Text.Collector, Text.TFIDF.
  • New function hanaml.OnlineLinearRegression(), an online version of linear regression.
  • New function hanaml.OnlineMultiLogisticRegression(), an online version of multi logistic regression.
  • New Bayesian Change-Point Detection function hanaml.BCPD().

Enhancement:

  • Categorical feature support in AdditiveModelForecast().
  • Supports SAP HANA Data Lake in ModelStorage.

API Changes:

  • Added a new parameter categorical.variable in hanaml.AdditiveModelForecast().

What’s New in hana.ml.r v2.9.210619

Bug Fixed:

  • Fixed missing default value of parameter ‘cols’ in hanaml.CovarianceMatrix and hanaml.PearsonrMatrix.
  • Fixed missing schema setting issue in hanaml.ConvertToHANADataFrame using JDBC methods to write the HANA table.
  • Fixed lowercase error of ACCURACY_MEASURE spelling.
  • Fixed the corrected the value of param intercept in linear regression.

What’s New in hana.ml.r v2.8.210321

Version 2.8.210321 supports SAP HANA SPS05 and SAP HANA Cloud

New Functions:

  • Added tbl() to support dbplyr
  • Added hana2dbplyr() and dbplyr2hana() to subpport dbplyr
  • New Unified functions: hanaml.UnifiedRegression, hanaml.UnifiedClustering.
  • New Markov Chain Monte Carlo function: hanaml.mcmc.
  • New ‘hana.version’ function is provided by hanaml.ConnectionContext.
  • New function hanaml.OnlineARIMA.
  • New function hanaml.VectorARIMA.
  • New function hanaml.sqltrace.

API Changes:

  • hanaml.KMeans with two added parameters ‘use.fast.library’ and ‘use.float’.
  • Added a parameter ‘distance.level’ in hanaml.UnifiedClustering when ‘func’ is AgglomerateHierarchicalClustering and DBSCAN. Please refer to documentation for details.
  • Added a parameter ‘range.penalty’ in hanaml.CPD.
  • Provides more methods of ‘decomposition’ in various regression functions, please refer to the documentation of the specific algorithm.
  • Added a parameter ‘key’ in prediction() function of hanaml.ARIMA and hanaml.AutoARIMA.
  • Added 2 parameters ‘min.measure’, ‘max.consequent’ in hanaml.KORD.
  • Added a parameter ‘output.threshold’ in hanaml.AUC to enable the output of threshold values in roc table.
  • Removed RODBC dependency.

Enhancement:

  • Improved the data upload in ConvertToHANADataFrame() for odbc and jdbc connection.
  • Enhanced hanaml.UnifiedClustering to support ‘distance.level’ in AgglomerateHierarchicalClustering and DBSCAN functions. Please refer to the documentation for details.
  • Enhanced the speed of data import functionality when odbc is applied.
  • Enhanced the validation of types of parameters.
  • Enhanced the parameter check of season.start and allow.linear in hanaml.AutoARIMA.
  • Enhanced the model storage for hanaml.ARIMA, hanaml.AutoARIMA, hanaml.VectorARIMA and hanaml.OnlineARIMA.

Bug Fixed:

  • Fixed the displacement of parameter ‘dispersion’ in hanaml.CPD.
  • Fixed the displacement of parameter ‘category.weight’ in hanaml.GaussianMixture.

What’s New in hana.ml.r v2.6.201016

API Changes:

  • HybridGradientBoostingClassifier, HybridGradientBoostingRegressor: added a parameter ‘adopt_prior’ to indicate whether to adopt the prior distribution as the initial point.
  • LinearRegression: added a parameter ‘features.must.select’ to specifies the column name that needs to be included in the final training model when executing the variable selection.

What’s New in hana.ml.r v2.6.200928

API Changes:

  • Added encrypt, validateCertificate and autocommit options to JDBC ConnectionContext.
  • SVC, SVR, OneClassSVM, SVRanking: added parameters ‘compression’, ‘max.bits’, ‘max.quantization.iter’ for model compression.
  • RandomForestClassifier: added parameters ‘compression’, ‘max.bits’, ‘quantize.rate’ for model compression.
  • RandomForestRegressor: added parameters ‘compression’, ‘max.bits’, ‘quantize.rate’, ‘fittings.quantization’ for model compression.
  • In prediction function ARIMA and AutoARIMA, new value ‘truncation.algorithm’ of forecast_method is introduced to improve the prediction performance.
  • New parameters ‘string.variable’, ‘variable.weight’ added in KNNClassifier, KNNRegressor and DBSCAN to enable distance calculation based on String distance.
  • New parameters ‘extrapolation’, ‘smooth.width’, ‘auxiliary.normalitytest’ are added in SeasonalDecompose.

New Functions:

  • Clustering: SlightSilhouette.
  • Preprocessing : SMOTETomek, TomekLinks.

Bug Fixed:

  • Fixed version check to support HANA cloud.
  • Fixed the error in random distribution sampling when full distribution parameter set is specified.
  • Fixed HAS_ID error in TomekLinks.
  • Fixed the Collect() method that returns ‘No Data’ when contraining a column of type ST_GEOMETRY(in particular, for GeoDBSCAN).

What’s New in hana.ml.r v2.5.200626

API Changes:

  • Removed parameter ConnectionContext in functions.
  • Added a new parameter ‘decay’ to replace ‘learning.rate’ as the later one is misleading in SOM().
  • Added a new parameter ‘stratified.columns’ in hanaml.Sampling() to replace ‘features’ whose name is misleading.
  • Added a new parameter ‘col.types’ in ConvertToHANADataFrame() to replace ‘clob.columns’ for enhancement.
  • Changed parameter ‘seed’ to ‘random.state’ for parameter name consistency in MLPClassifier() and MLPRegressor().
  • Update function names for consistency: Arima -> ARIMA, AutoArima -> AutoARIMA， Auc -> AUC, Kmeans -> KMeans, Kmedian -> KMedian, Kmedoid -> KMedoid, Kord -> KORD.

New Functions:

  • Recommender System Algorithms : Alternating Least Square, Factorized Polynomial Regression Models, Field-Aware Factorization Machine.
  • Regression : Cox Proportional Hazard Model.
  • Statistics Functions : Cumulative Distribution Function (CDF), Distribution Fitting, Distribution Quantile, Entropy, Equal Variance Test, Factor Analysis, Wilcoxon Signed Rank Test, Grubbs’ Test, Kaplan-Meier Survival Analysis, Kernel Density Estimation, One-Sample Median Test.
  • Time Series : Fast DTW
  • Preprocessing : SMOTE
  • Miscellaneous : ABC Analysis, T-Distributed Stochastic Neighbour Embedding(TSNE), Weighted Score Table.
  • Unified classification
  • Model Selection

Bug Fixed:

  • Fixed parameter ‘formula’ parsing issue when a single feature is entered.
  • Fixed falsely recognizing the type of query statement with JDBC in sqlQueryMix().
  • Fixed phrasing error of parameter ‘timeout’ in Decision Tree, Linear Regression, Logistic Regression, Naive Bayes and SVM functions.

Enhancement:

  • Added cross-validation options to some functions (Decision Tree Classifier/Regressor, Gradient Boosting Classifier/Regressor, Hybrid Gradient Boosting Classifier/Regressor, Generalised Linear Models(GLM), Naive Bayes, Linear Regression, Logistic Regression Multi-Layer Perceptron Classifier/Regressor, Support Vector Machines functions, K-Nearest Neighbors Classifier/Regressor, Alternating Least Square(ALS), Factorized Polynomial Regression Models(FRM), Polynomial Regression).
  • Improved the robustness of ConvertToHANADataFrame().
  • Enhancement of ConvertToHANADataFrame() function: support data.frame input with missing (NA) values.

What’s New in hana.ml.r v1.0.8

Bug Fixed:

  • Fixed wrong error message with RJDBC connection. Add two additional error message for missing RJDBC connection and wrong RJDBC connection.context type.
  • Fix label type error in RandomForest. Add process to check continuous label type for regression and categorical label type for classification.
  • Fixed wrong API for DecisionTreeRegressor. Remove type conversion for result.
  • Fixed wrong cast error in Neighbors. Use NVARCHAR instead of DOUBLE for result.
  • Fixed wrong cast error in HGBT. Use NVARCHAR instead of DOUBLE for result.

What’s New in hana.ml.r v1.0.7

New Algorithms:

  • Association : Apriori, Apriorilite, FP-Growth, KORD, Sequential Pattern Mining (SPM)
  • Clustering : Affinity Propagation, Agglomerate Hierarchical Clustering, DBSCAN, Geometry DBSCAN, Latent Dirichlet Allocation, Self-Organizing Maps (SOM)
  • Classification : Conditional Random Field (CRF), Confusion Matrix, Hybrid Gradient Boosting Tree (HGBT), Logistic Regression, Multilayer Perceptron (MLP).
  • Regression : Bi-Variate Geometric Regression, Bi-Variate Natural Logarithmic Regression, Exponential Regression.
  • Time Series : ARIMA, Auto ARIMA, FFT, Seasonal Decompose, Trend Test, White Noise Test Single/Double/Triple/Auto/Brown Exponential Smoothing, Change-Point Detection, Croston’s Method, Linear Regression With Damped Trend And Seasonal Adjust, Additive Model Forecast, Hierarchical Forecast, Correlation Function.
  • Preprocessing : Discretize, Inter-Quartile Range (IQR), Missing Value Handing, Multidimensional Scaling (MDS), Partition, Random Distribution Sampling, Variance Test function.
  • Statistics Functions ： Chi-Squared Test Functions, T-Test Functions, Analysis Of Variance Functions (ANOVA), Univariate/Multivariate Analysis Functions.
  • Random Distribution Sampling Functions ： Bernoulli, Beta, Binomial, Cauchy, Chi_Squared, Exponential, Extreme_Value, F, Gamma, Geometric, Gumbel, Lognormal, Negative_Binomial, Normal, Pert, Poisson, Student_T, Uniform, Weibull, Multinomial).
  • Social Networks : Link Prediction, Pagerank.
  • Connection Context : JDBC Option.
  • Model Storage Services.
  • Dataframe Functions : ConvertToHANADataFrame.

Quality Improvements:

  • Use Anonymous Block.
  • Remove Unnecessary Temporary Table.
  • Error Message Enhancement.
  • Fix Documentation.
  • Code Quality Improvement.