Spaces:
Sleeping
Sleeping
Atharva Thakur
commited on
Commit
•
6b53acf
1
Parent(s):
d6e847d
Update README.md
Browse files
README.md
CHANGED
@@ -13,25 +13,96 @@ pinned: false
|
|
13 |
## Deployment
|
14 |
[HuggingFace](https://huggingface.co/spaces/AtharvaThakur/Insights)
|
15 |
|
16 |
-
|
17 |
|
18 |
-
|
19 |
-
|
20 |
-
-
|
21 |
-
|
22 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
|
24 |
## Features
|
25 |
|
26 |
-
-
|
27 |
-
-
|
28 |
-
-
|
29 |
-
-
|
30 |
-
-
|
31 |
-
-
|
32 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
35 |
|
36 |
1. Install the required packages:
|
37 |
The project's dependencies are listed in the 'requirements.txt' file. You can install all of them using pip:
|
@@ -44,28 +115,6 @@ pinned: false
|
|
44 |
streamlit run app.py
|
45 |
```
|
46 |
|
47 |
-
## Web app
|
48 |
-
1. Main page
|
49 |
-
Data Exploration
|
50 |
-
-> Data Loader
|
51 |
-
-> DataQA (LLM with python interpreter/CSV agent)
|
52 |
-
-> Data Analyzer
|
53 |
-
-> Data Filter
|
54 |
-
-> Data Visualizer
|
55 |
-
|
56 |
-
2. Data Transformation
|
57 |
-
-> handling null values
|
58 |
-
-> creating new columns
|
59 |
-
-> removing columns
|
60 |
-
-> Changing datatypes
|
61 |
-
-> give option to analyse the transformed dataset or save it.
|
62 |
-
|
63 |
-
3. Natural language dataparty (Pure LLM)
|
64 |
-
-> Insights generation
|
65 |
-
-> Automating the data analysis/transformation
|
66 |
-
-> generating a report
|
67 |
-
|
68 |
-
|
69 |
# Running using Docker
|
70 |
1. Build the docker image using
|
71 |
```
|
|
|
13 |
## Deployment
|
14 |
[HuggingFace](https://huggingface.co/spaces/AtharvaThakur/Insights)
|
15 |
|
16 |
+
# Insights: Gen-AI Based Data Analysis Tool
|
17 |
|
18 |
+
## Overview
|
19 |
+
|
20 |
+
**Insights** is a state-of-the-art data analysis tool that leverages the Gemini-Pro large language model (LLM) to automate and enhance the data analysis process. This tool aims to perform end-to-end data analysis tasks, providing substantial cost and time savings while matching or exceeding the performance of junior data analysts.
|
21 |
+
|
22 |
+
## Table of Contents
|
23 |
+
|
24 |
+
1. [Introduction](#introduction)
|
25 |
+
2. [Features](#features)
|
26 |
+
3. [System Architecture](#system-architecture)
|
27 |
+
4. [Modules Overview](#modules-overview)
|
28 |
+
5. [Installation](#installation)
|
29 |
+
6. [Usage](#usage)
|
30 |
+
7. [Evaluation](#evaluation)
|
31 |
+
8. [Contributors](#contributors)
|
32 |
+
9. [License](#license)
|
33 |
+
|
34 |
+
## Introduction
|
35 |
+
|
36 |
+
In today's data-driven world, robust data analysis tools are crucial for informed decision-making and strategic planning. Traditional data analysis methods often face challenges such as time-consuming processes, potential for errors, and the need for specialized expertise. **Insights** addresses these issues by utilizing AI to streamline and enhance the data analysis process.
|
37 |
|
38 |
## Features
|
39 |
|
40 |
+
- **Automated Data Analysis**: Perform data collection, visualization, and analysis with minimal human intervention.
|
41 |
+
- **Advanced Summarization**: Generate detailed summaries and potential questions for datasets.
|
42 |
+
- **Exploratory Data Analysis (EDA)**: Tools for statistical summaries, distribution plots, and correlation matrices.
|
43 |
+
- **Data Cleaning and Transformation**: Functions for handling missing values, outlier detection, normalization, and feature engineering.
|
44 |
+
- **Machine Learning Toolkit**: Automates model selection, training, hyperparameter tuning, and evaluation.
|
45 |
+
- **Query Answering Module**: Generate Python code to answer user queries and produce visualizations.
|
46 |
+
|
47 |
+
## System Architecture
|
48 |
+
|
49 |
+
The **Insights** tool is built on the Gemini platform and consists of three main components:
|
50 |
+
|
51 |
+
1. **Summary Module**
|
52 |
+
2. **QA Module**
|
53 |
+
3. **Code Execution and Analysis Generation**
|
54 |
+
|
55 |
+
### Summary Module
|
56 |
+
|
57 |
+
Extracts essential details about the dataset and generates a comprehensive summary along with potential questions for further exploration.
|
58 |
+
|
59 |
+
### QA Module
|
60 |
+
|
61 |
+
Handles user queries related to the dataset, generating Python code to answer the queries and produce visualizations.
|
62 |
+
|
63 |
+
### Code Execution and Analysis Generation
|
64 |
+
|
65 |
+
Executes the generated Python code offline to ensure data security, producing detailed responses and visualizations.
|
66 |
+
|
67 |
+
## Modules Overview
|
68 |
|
69 |
+
### Summary Generation
|
70 |
+
|
71 |
+
1. **Information Extraction**: Extracts critical information from the dataset.
|
72 |
+
2. **Prompting Gemini**: Constructs a detailed prompt for Gemini to generate summaries and questions.
|
73 |
+
3. **Summary and Question Generation**: Generates a summary and potential questions for user review.
|
74 |
+
|
75 |
+
### Data Exploration
|
76 |
+
|
77 |
+
Includes tools for EDA, data cleaning, and data transformation.
|
78 |
+
|
79 |
+
### ML Toolkit
|
80 |
+
|
81 |
+
Facilitates the creation and evaluation of machine learning models on the dataset.
|
82 |
+
|
83 |
+
### QA Module
|
84 |
+
|
85 |
+
Allows users to query the dataset and receive answers along with visualizations. The process involves:
|
86 |
+
|
87 |
+
1. Accepting user queries.
|
88 |
+
2. Combining queries with dataset information.
|
89 |
+
3. Generating and executing Python code offline.
|
90 |
+
4. Producing visualizations and textual data.
|
91 |
+
|
92 |
+
### Analysis Generation
|
93 |
+
|
94 |
+
Processes the output from code execution to create concise and insightful responses.
|
95 |
+
|
96 |
+
## Usage
|
97 |
+
1. Initialize the Tool:
|
98 |
+
`python app.py`
|
99 |
+
2. Load Dataset: Upload your dataset when prompted.
|
100 |
+
3. Generate Summary: The tool will automatically generate a summary and potential questions.
|
101 |
+
4. Exploratory Data Analysis: Use the EDA tools to explore your dataset.
|
102 |
+
5. Query the Dataset: Enter your queries to receive answers and visualizations.
|
103 |
+
6. Analyze Results: Review the detailed analysis generated by the tool.
|
104 |
+
|
105 |
+
## Installation Instructions
|
106 |
|
107 |
1. Install the required packages:
|
108 |
The project's dependencies are listed in the 'requirements.txt' file. You can install all of them using pip:
|
|
|
115 |
streamlit run app.py
|
116 |
```
|
117 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
118 |
# Running using Docker
|
119 |
1. Build the docker image using
|
120 |
```
|