What Data Scientists Chose in 2018: Top AI, ML, and Big Data Tools Revealed

Survey Overview

In 2018, the well‑known data‑science site KDnuggets released the results of its annual software survey, gathering responses from more than 2,000 participants about the data‑mining and machine‑learning tools and programming languages they used in the past twelve months.

The survey also compared rankings with the previous three years and excluded "single‑tool" votes that could skew the results.

Key Findings

Programming Languages

Python’s share rose to 65.6% (up 11% from 2017), while R fell to 48.5% (down 14%). SQL remained steady at about 40%, and other languages such as Java, C/C++, and Scala showed modest declines.

Deep‑Learning Frameworks

TensorFlow remained the most popular deep‑learning framework with 29.9% usage, followed by Keras at 22.2% and PyTorch at 6.4%.

Tool Popularity

RapidMiner’s usage jumped from 33% in 2017 to 52.7% in 2018, making it the fastest‑growing data‑science platform. Hadoop’s share dropped by more than 30%, while Spark‑SQL appeared as the only new tool surpassing a 2% usage threshold (11.7%).

Stability and Decline

SQL (including Spark SQL and Hadoop‑related tools) held a stable ~40% share. Conversely, many established tools saw usage declines of 20% or more, as shown in the “largest drop” tables.

Regional Participation

Europe contributed 37.5% of responses (up from 35.5%), while North America fell to 36.6% (down from 41.5%). Other regions included Asia (11.7%), Latin America (6.6%), Africa/Middle East (4.5%), and Australia/New Zealand (3.1%).

Visual Data

Source: Machine Heart (机器之心) – original article linked in the post.