↓Skip to main content

Analysing and automating service requests

·201 words·1 min
Scikit Learn Clustering KMeans DBScan Explorative Data Analysis (EDA) LLM Streamlit Plotly Machine Learning Data Visualisation Hyperparameter Optimisation Huggingface Text Embedding PCA Classification Python Azure Cloud Data Scientist Git Confluence Jira

The project involves analyzing and automating service requests. For this purpose, 3.5 years of data are exported from an issue tracking tool and analyzed.

Exploratory data analysis #

Firstly, an exploratory data analysis (EDA) is carried out to gain a better understanding of the data structure and distribution. Basic statistical key figures are calculated and visualizations are created.

The texts of the service requests are represented in high-dimensional vectors using Huggingface SentenceTransformer Models. These embeddings serve as the basis for clustering and further classification.

Clustering methods such as k-Means and DBSCAN are used to identify patterns in the service requests. Before clustering, dimension reduction is performed using Principal Component Analysis (PCA). The resulting clusters are visualized to understand the groupings of the requests.

Automatic classification #

Based on the determined clusters and the generated embeddings, models are created for the automatic classification of the service requests. Various machine learning approaches are used to automatically assign requests to the appropriate categories.

Want to learn more? Make an appointment now!

Activities #

  • Execution of Explorative Data Analysis (EDA)
  • Generation of data visualisations
  • Calculation of clusters with k-Means and DBSCAN
  • Transfer to vector space with Huggingface Transformer models
  • Classification with various ML models (e.g. RandomForest, XGBoost)