How Sugar BI Merges AI, Big Data, and Zero‑Code Design for Smarter Visual Analytics

1. Sugar BI Introduction

Sugar BI is presented as a data‑value mining platform that directly connects data with users, offering a professional, scenario‑driven BI analysis environment.

1.1 Baidu Intelligent Cloud Big Data System Architecture Overview

The Baidu Intelligent Cloud big‑data product architecture consists of three layers:

Bottom layer: Lake‑warehouse infrastructure providing storage, processing, and development capabilities.

Middle layer: Data value mining platform that maximizes enterprise data asset value.

Top layer: Industry‑specific big‑data application deployment.

A data security protection system (multi‑party computation, data audit, encryption, desensitization) is also integrated.

1.2 Rapid Professional BI Platform Construction

Goal: enable users to build a professional, scenario‑based BI platform within five minutes.

Process: add data source → create data model → design visualizations (reports & dashboards) using Apache ECharts drag‑and‑drop components.

1.3 Multi‑Source Data Integration

Supported data sources include open relational databases (MySQL, SQL Server, PostgreSQL, Oracle), big‑data sources (Kylin, Hive, Spark, Impala, Presto), file uploads (Excel/CSV), API connections, and JSON static inputs. An internal tunnel feature allows on‑premise data sources to connect securely to the cloud service, and cross‑source analysis is supported.

1.4 Zero‑Code Drag‑Drop Report Creation

All components on the report editor are drag‑and‑drop.

Fields are bound by dragging, enabling full‑screen chart addition and modification.

Rich chart configuration allows users to start without any coding.

Responsive mobile layout adapts automatically.

Empowers business users to perform analysis directly.

2. Visualization Technology Analysis

2.1 Sugar BI Architecture – AI‑Featured Visualization + BI Platform

Sugar BI is positioned as a visualization analysis platform with three core characteristics: AI, BI, and visual rendering. The architecture fuses these capabilities and is containerized for deployment.

2.2 BI Capabilities – Data Model, Flexible Interaction, Data Computation

Supports over 30 data sources, including MySQL, Oracle, Doris, ClickHouse, and domestic databases such as Kylin, GaussDB, etc.

Data Model: Enables single‑table queries, multi‑table joins, view usage, and automatic dimension/measure detection during drag‑and‑drop.

BI Engine: Parses user interactions to generate filter conditions, drill‑downs, and URL‑linked queries that run directly on the database.

Computation Engine: Provides table calculations, cross‑calculations, formatting, monitoring, and retention analysis.

2.2.1 Data Model Advantages

Supports open databases (MySQL, SQL Server, PostgreSQL, Oracle).

Zero‑code: users need not write SQL.

Abstracts SQL differences across sources.

Automatically distinguishes dimensions, measures, and geographic fields based on type and naming.

Standard aggregations (SUM, AVG, MAX, MIN) are available.

2.2.2 Flexible Interaction

Complex WHERE filtering is handled via a left‑side UI that supports intricate conditions, including HAVING after aggregation. Both page‑level and chart‑level filter components are provided, and chart‑to‑chart linking enables coordinated drill‑downs.

2.2.3 SQL‑Level Data Computation

Data format conversion

Date‑time aggregation

Calculated fields

Bucket grouping

2.2.3.1 Secondary Computation

After SQL execution, results are processed in memory using multithreading for value mapping, cross‑pivot tables, totals, averages, and fast table calculations.

2.2.3.2 Performance Optimization

Result caching via Redis to avoid repeated SQL execution.

Source‑specific optimizations (e.g., ClickHouse retention functions).

Asynchronous multithreaded table calculations.

3. Intelligent Chart Recommendation

3.1 Purpose

Automatically suggests the most suitable chart type based on data dimensions, measures, and geographic fields, while allowing manual switching.

3.2 Recommendation Process

Features of over 100 chart types are abstracted into characteristics; user‑dragged fields are transformed into data features, matched against chart features, scored, and the highest‑scoring chart is recommended.

3.3 Chart Feature Design

Defines required dimensions, measures, and field limits for each chart, handling variations such as stacked vs. single lines and mixed chart types.

3.4 Recommendation Strategy

Combines mandatory rules (must be satisfied, otherwise score 0) with optional rules (additive scoring). Mandatory rules ensure every field appears in the recommended chart; optional rules consider similarity of field characteristics, date relevance, and measure units.

4. Intelligent Voice Interaction

4.1 Voice Q&A Overall Scheme

The Q&A model leverages the data model’s synonyms, recommended questions, dimensions, and schema. Natural Language Understanding (NLU) interprets user queries, maps them to fields, applies filters, sorting, and aggregation, and finally renders the result via the smart chart engine.

5. Future Plans and BI Trends

5.1 BI Analysis Trends

Attribution analysis – automatically identifies dimensions driving metric changes (already launched).

Anomaly analysis – detects sudden data shifts and issues alerts (already launched).

Predictive analysis – applies machine‑learning/AI to forecast future trends.

Ba​idu Intelligent Cloud believes the next wave of BI is the fusion of AI and BI capabilities.