Patent Analysis

This is a tool to download US Patent information, particularly, citation network from patentsview.org and store into MongoDB to analyze. Currently, this dataset contains 6.215.171 patents (verticies) and 86.184.397 citations (edges).

For example, we can draw in-degree distribution of the citation network.

or average citations (out-degree) added per year:

or edge/node (loglog) count over time:

See the notebook for other demos.

Getting Started

1) Python dependencies

First, you need to have the required packages under requirements.txt. You can install using virtualenv or to user.

2) Database

Second, you need a MongoDB running. If you have docker you can use following line to run an instance:

$ docker run --name patent-mongo -v `pwd`/db:/data/db -p 27017:27017 -d mongo:3.4.6

(If your db host is not localhost, set it with environment vars like $ export MONGO_URL="mongodb://host:27017" before running the app.)

3) Downloading and storing the dataset:

This will download the dataset from patentsview.org/download/:

$ ./download_data.sh

Now inserting it to the db:

$ python -m patent_analysis.insert_db

(It takes ~4h with SSD Disk!)

Now we are ready to work on the patent data.

Roadmap

• Queries for WIPO categories
• Centrality measure analysis such as PageRank
• Community analysis
• Embeddings
• Assignee network analysis
• ...

License

Released under the MIT license.