R leads RapidMiner, Python catches up, Big Data tools grow, Spark ignites


R is the most popular overall tool among data miners, although Python usage is growing faster. RapidMiner continues to be most popular suite for data mining/data science. Hadoop/Big Data tools usage grew to 29%, propelled by 3x growth in Spark. Other tools with strong growth include H2O (0xdata), Actian, MLlib, and Alteryx.





The 16th annual KDnuggets Software Poll continued to get huge attention from analytics and data mining community and vendors, attracting about 2,800 voters, who chose from a record number of 93 different tools.

R is the most popular overall tool among data miners and data scientists, but Python usage grows faster and it is likely to catch up in 2-3 years. RapidMiner remains the most popular suite for data mining/data science, but it got fewer votes than last year. There was a notable increase in Hadoop/Big Data tool usage (29%, up from 17% in 2014), mainly driven by jump in Spark whose usage share grew over 3-fold. (see KDnuggets exclusive interview with Spark Creator Matei Zaharia). Other tools with strong growth include H2O (0xdata), Actian, MLlib, and Alteryx.

This report has 5 sections
 
The participation by region was: US/Canada (41.5%), Europe (38.4%), Asia (8.2%), Latin America (6.3%), Australia/NZ (3.1%), Africa/MidEast (2.5%).
 

Top Analytics Tools and Trends

Here are the top 10 tools by share of usage:

Top10 Analytics Data Mining Software 2015

The top 10 tools by share of users were
  1. R, 46.9% share ( 38.5% in 2014)
  2. RapidMiner, 31.5% ( 44.2% in 2014)
  3. SQL, 30.9% ( 25.3% in 2014)
  4. Python, 30.3% ( 19.5% in 2014)
  5. Excel, 22.9% ( 25.8% in 2014)
  6. KNIME, 20.0% ( 15.0% in 2014)
  7. Hadoop, 18.4% ( 12.7% in 2014)
  8. Tableau, 12.4% ( 9.1% in 2014)
  9. SAS, 11.3 (10.9% in 2014)
  10. Spark, 11.3% ( 2.6% in 2014)

 
Compared to 2014 Analytics/Data Mining Software Poll, Tableau and Spark were newcomers to top 10, displacing Weka and Microsoft SQL Server.

The average number of tools jumped to 4.8, up from 3.7 in 2014 and 3.0 in 2013.

The distinction between commercial and free software is becoming harder to make, with many tools having both a free/community version and commercial/enterprise version. We classified each tool according to the primary type of the latest version, so we put RapidMiner in commercial category and KNIME in free software category.

Many vendors asked their users to vote in the poll and even tweet their vote, but we have not found any bot or illegal voting, and did not have to remove any votes.

This year, 91% of voters used commercial software and 73% used free software. About 27% used only commercial software, and only 9% used free-software. For the first time a majority of 64% used both free and commercial software, up from 49% in 2014.

Analytics Data Mining Software Commercial Free Venn

Among tools with at least 10 votes, the highest increase in 2014 was for
  1. H2O (0xdata), 1210% up, to 2.0% share (55 votes) from 0.2% in 2014
  2. Actian, 345% up, to 2.0% (56 votes), from 0.5% in 2014
  3. Spark, 326% up, to 11.3% (311), from 2.6% in 2014
  4. MLlib, 228% up, to 3.3% (91), from 1.0% in 2014
  5. Alteryx, 79% up, to 5.6% (155), from 3.1% in 2014
  6. Python, 56% up, to 30.3% (837), from 19.5% in 2014
  7. TIBCO Spotfire, 56% up, to 4.3% (119), from 2.8% in 2014
  8. Pig, 54% up, to 5.4% (150), from 3.5% in 2014
  9. SAS Enterprise Miner, 53% up, to 10.9% (302), from 7.2% in 2014
  10. Splunk/Hunk, 49% up, to 1.1% (30), from 0.7% in 2014


Tools that showed at least 20% increases in their share for 2 years in the row are Alteryx, Hadoop, KNIME, Python, Qlikview, SAS Enterprise Miner, Tableau, and TIBCO Spotfire.

New analytics tools that received at least 20 votes in 2015 were
  • scikit-learn, 8.3% (229)
  • Microsoft Azure ML, 3.7% (102)
  • Microsoft Power BI, 3.6% (98)
  • IBM Watson Analytics, 2.1% (57)
  • Ayasdi, 2.0% (56)
  • Dataiku, 2.0% (56)
  • Lexalytics, 1.3% (35)
  • Vowpal Wabbit, 1.3% (35)
  • Microstrategy, 0.9% (24)
  • Amazon Machine Learning, 0.7% (20)

 


Among tools with at least 20 votes in 2014, the largest decline in 2015 was for these tools, which includes probably a combination of decline of popularity for free tools like Orange and lack of a voter drive for some of commercial tools this year.
  • Predixion Software, 90% down (0.4% share), from 3.7% in 2014
  • BayesiaLab, 86% down, to 0.6%, from 4.1%
  • Alpine Data Labs, 82% down, to 0.5% from 2.7%
  • Oracle Data Miner, 64% down, to 0.8% from 2.2%
  • RapidInsight/Veera, 60% down, to 0.2% from 0.5%
  • Revolution Analytics (now part of Microsoft), 57% down, to 4.0% from 9.1%
  • SAP (including former KXEN), 57% down, to 3.0% from 6.8%
  • Orange, 44% down to 1.9% from 3.4%
  • Gnu Octave, 41% down, to 2.3% from 3.9%

 


Hadoop/Big Data Tools

Hadoop/Big Data tool usage jumped to 29% among voters, up from 17% in 2014, and 14% in 2013.

This is probably due to availability and low-cost of many cloud-based Big Data tools. Very notable is the jump in Spark share to 11.3%.

However, most data analysis is still done on "medium" and small data.

Top Hadoop/Big Data tools were
  • Hadoop, 18.4% share (507 votes)
  • Spark, 11.3% (311)
  • Hive, 10.2% (282)
  • SQL on Hadoop tools, 7.2% (198)
  • Pig, 5.4% (150)
  • HBase, 4.6% (127)
  • Other Hadoop/HDFS-based tools, 4.5% (125)
  • MLlib, 3.3% (91)
  • Mahout, 2.8% (76)
  • Datameer, 0.8% (23)

 

Deep Learning Tools

New this year was a category of Deep Learning Tools, with most popular tools being:
  • Pylearn2 (55 users)
  • Theano (50)
  • Caffe (29)
  • Cuda-convnet (17)
  • Deeplearning4j (12)
  • Torch (27)

 

However, this category is growing rapidly and above list is incomplete, since the largest count in this category was for other Deep Learning tools (106)

See also
 

Programming Languages

Python increased significantly in popularity. Java is the second most commonly used language for analytics/data mining tasks. Here is the
  • Python, 30.3% share (837 votes), up from 19.5%
  • Java, 14.2% (392), na in 2014
  • C/C++, 9.4% (260), na in 2014
  • Unix shell/awk/gawk, 8.0% (221), up from 5.8%
  • Other programming languages, 5.1% (140)
  • Scala, 3.5% (96), na in 2014
  • Perl, 2.9% (79), down from 3.0
  • Ruby, 1.2% (33), na in 2014
  • Julia, 1.1% (31), up from 0.8%
  • F#, 0.7% (18), up from 0.5%
  • Clojure, 0.5% (13), same as 0.5%
  • Lisp, 0.4% (10), up from 0.3%

 

Full Results and 3-year trends

The following table shows the poll results in detail.
% alone is the percent of tool voters used only that tool alone. For example, only 3.6% of R users have used only R, while 13.7% of RapidMiner users indicated they used that tool alone.

What Analytics, Big Data, Data mining, Data Science software you used in the past 12 months for a real project? [2759 voters]
Legend: Red: Free/Open Source tools
Green: Commercial tools
Fuchsia: Hadoop/Big Data tools
% users in 2015
% users in 2014
% users in 2013
R (1293), 3.6% alone
RapidMiner (870), 13.7% alone
SQL (853), 0% alone na
Python (837), 0% alone
Excel (631), 0% alone
KNIME (553), 6.7% alone
Hadoop (507), 0% alone
Tableau (341), 0% alone
SAS base (313), 0.6% alone
Spark (311), 0% alone na
Weka (310), 0% alone
SAS Enterprise Miner (302), 3.6% alone
Microsoft SQL Server (268), 0% alone
MATLAB (243), 0% alone
scikit-learn (229), 0% alone na
na
Unix shell/awk/gawk (221), 0% alone na
IBM SPSS Statistics (213), 0% alone
IBM SPSS Modeler (197), 7.1% alone
Alteryx (155), 39.4% alone
Pig (150), 0% alone na
Other programming languages (140), 0% alone na
Other free analytics/data mining tools (138), 0% alone
Other Hadoop/HDFS-based tools (125), 0% alone na
TIBCO Spotfire (119), 11.8% alone
Rattle (117), 0.9% alone
QlikView (116), 0% alone
Revolution Analytics (now part of Microsoft) (109), 0% alone
Microsoft Azure ML (102), 1.0% alone na
na
Microsoft Power BI (98), 0% alone na
na
MLlib (91), 0% alone na
JMP (86), 0% alone
SAP (including former KXEN) (82), 26.8% alone
Perl (79), 0% alone na
Mahout (76), 0% alone na
Pentaho (74), 0% alone na
Other paid analytics/data mining/data science software (66), 6.1% alone
Salford SPM/CART/Random Forests/MARS/TreeNet (64), 43.8% alone
Gnu Octave (64), 0% alone
IBM Watson Analytics (57), 0% alone na
na
Ayasdi (56), 10.7% alone na
na
Dataiku (56), 7.1% alone na
na
Actian (56), 7.1% alone na
H2O (0xdata) (55), 0% alone na
Orange (53), 0% alone
Mathematica (52), 0% alone
IBM Cognos (51), 0% alone
Dell (including StatSoft) (47), 19.1% alone
XLSTAT for Excel (42), 0% alone
Stata (36), 2.8% alone
Lexalytics (35), 28.6% alone na
na
Vowpal Wabbit (35), 0% alone na
na
C4.5/C5.0/See5 (35), 0% alone
Julia (31), 3.2% alone na
Splunk/ Hunk (30), 0% alone na
Datameer (26), 0% alone na
MicroStrategy (24), 0% alone na
na
BigML (23), 0% alone na
Zementis (22), 31.8% alone
Miner3D (22), 9.1% alone
Oracle Data Miner (22), 0% alone
Amazon Machine Learning (20), 5.0% alone na
na
F# (18), 0% alone
BayesiaLab (16), 12.5% alone
Dato (former Graphlab) (15), 6.7% alone na
Clojure (13), 0% alone na
Alpine Data Labs (13), 0% alone na
Angoss (11), 18.2% alone
Lavastorm (10), 0% alone
Lisp (10), 0% alone na
Predixion Software (10), 0% alone
WordStat (9), 0% alone
Megaputer Polyanalyst/TextAnalyst (8), 0% alone
WPS: World Programming System (7), 0% alone na
GoodData (6), 0% alone na
MetaMind (5), 0% alone na
na
SiSense (5), 0% alone na
RapidInsight/Veera (5), 0% alone
Skytree (3), 0% alone na
na
Birst (2), 0% alone na
na
Ontotext (1), 0% alone na
na
FICO Model Builder (1), 0% alone na


Additional tools not included but mentioned in the comments include
  • Daniel Soto: ETL: Anatella; predictive analytics: TIMI modeler.
  • Henrique Pinto: proposed separation of SAP technologies into the modeling tool (SAP Predictive Analytics, which merges SAP PA + KXEN) and SAP HANA as the underlying platform, in the same sense you have SAS Miner and SAS Base. HANA has its own programming logic (based on SQL, called SQLScript) which can be used for native development of predictive models, or you could use SAP Predictive Analytics high-level modeling capabilities on top of HANA for less development capable users.
  • Another tool suggestion: Domino (DominoLabs), analytical hub for sophisticated enterprises: helps organizations develop, track and deploy their analytical models faster, while facilitating best practices by keeping work centralized, sharable, and auditable.
  • Roberto Lopez: Neural Designer, a predictive analytics tool with high performance.
  • Julian GV: Experian Strategy Management, that includes Assisted Design, the analytics module integrated with the software. That's the solution I used in the past 12 months.
  • Universal Platform, UP

 
Here are the results of past polls