/Large-Language-Models

Large language models offer new opportunities for processing and generating text. I used text embeddings, clustering, and the ChatGPT API to examine the reasons for startup failure.

Primary LanguagePython

Dataset

CB Insights, 463 startup failure post-mortems from Aug 23, 2023

https://www.cbinsights.com/research/startup-failure-post-mortem/#2023updateQ223

I used the failure post-mortems for the 118 startups between 2021 and 2023. See the attached stories.csv file

#Prompt to extract failure reasons

Extract the failure reasons from this start-up shutdown. Each reason should be five words or less.

#Response from the Streamlit application

streamlit app1

#Dataframe showing the failure stories in the dataset

Dataframe

Elbow and Cluster

#Labels of the five largest clusters

The labeling was done based on the most prominent reason for failure common to the companies in each cluster.

Cluster 2: Financial Struggles

Cluster 6: Capital Drought

Cluster 19: Capital Struggle

Cluster 0: Strategic Adaptation Challenges

Cluster 3: Biotech Challenge

#Failure reason common to each cluster is 'Lack of Funding'

#Top 5 reasons for startup failure

(i) Lack of funding

(ii) Investors not found

(iii) No product/market fit

(iv) No/wrong business model

(v) Political/economic/legal problem