Junior DS 25.12.2022. Retail recommender system. User's traffic classification. Music recommendatio
Join us: https://discord.gg/sZkePhaWSZ ... --- The group is discussing new datasets related to recommendation systems and exploring different tasks such as predicting properties of events using visitor data and predicting categories and quantities of products. They are encountering difficulty in understanding some of the data and need further exploration. They suggest creating a legend for the "Property" column and grouping users based on their interactions with certain products to improve recommendation algorithms. They consider grouping users based on the number of times they view a particular product before purchasing it and creating a frequency chart to identify the most popular products. They also discuss the need for specific features to visualize the data and improve the accuracy of the predictions. Additionally, they discuss the importance of considering a user's history of purchases and interactions with specific products when making recommendations. They suggest grouping users based on the frequency of their interactions with certain products and using this information to assign weights to different products. They also discuss the need to analyze the likelihood of a user making a purchase after viewing a product and the importance of visualizing this data to improve the accuracy of the predictions. The group also discusses a complex method for analyzing a table related to the cost of an object and deciphering the values in the "value" column, which are hashed. They suggest using a list to aggregate the data and make it easier to understand. They also discuss the need to analyze a table related to the interactions between users and products, including views, transactions, and adding items to a cart. They consider grouping users based on their interactions with certain products and using this information to improve recommendation algorithms. They also discuss the need to accurately identify the product characteristics at the time of the event to improve the accuracy of the recommendations. They suggest freezing the data to ensure that the product characteristics match the event at the time it occurred. They also discuss the use of a suffix to identify changes in a product's characteristics over time and the need to match events with the corresponding product characteristics at the time of the event. They mention the use of weekly snapshots to record current product characteristics and the removal of values that do not change over time to reduce the size of the database. The group also discusses the importance of filtering data based on user actions such as registration and purchase and the need to limit the data to a maximum of 200 entries. They also mention the use of categories to improve the accuracy of the recommendations and the possibility of calculating ratings based on the frequency of interactions within a category. Additionally, they briefly discuss the use of neural networks for image classification and the process of passing image data and labels to the network.
Join us: https://discord.gg/sZkePhaWSZ ... --- The group is discussing new datasets related to recommendation systems and exploring different tasks such as predicting properties of events using visitor data and predicting categories and quantities of products. They are encountering difficulty in understanding some of the data and need further exploration. They suggest creating a legend for the "Property" column and grouping users based on their interactions with certain products to improve recommendation algorithms. They consider grouping users based on the number of times they view a particular product before purchasing it and creating a frequency chart to identify the most popular products. They also discuss the need for specific features to visualize the data and improve the accuracy of the predictions. Additionally, they discuss the importance of considering a user's history of purchases and interactions with specific products when making recommendations. They suggest grouping users based on the frequency of their interactions with certain products and using this information to assign weights to different products. They also discuss the need to analyze the likelihood of a user making a purchase after viewing a product and the importance of visualizing this data to improve the accuracy of the predictions. The group also discusses a complex method for analyzing a table related to the cost of an object and deciphering the values in the "value" column, which are hashed. They suggest using a list to aggregate the data and make it easier to understand. They also discuss the need to analyze a table related to the interactions between users and products, including views, transactions, and adding items to a cart. They consider grouping users based on their interactions with certain products and using this information to improve recommendation algorithms. They also discuss the need to accurately identify the product characteristics at the time of the event to improve the accuracy of the recommendations. They suggest freezing the data to ensure that the product characteristics match the event at the time it occurred. They also discuss the use of a suffix to identify changes in a product's characteristics over time and the need to match events with the corresponding product characteristics at the time of the event. They mention the use of weekly snapshots to record current product characteristics and the removal of values that do not change over time to reduce the size of the database. The group also discusses the importance of filtering data based on user actions such as registration and purchase and the need to limit the data to a maximum of 200 entries. They also mention the use of categories to improve the accuracy of the recommendations and the possibility of calculating ratings based on the frequency of interactions within a category. Additionally, they briefly discuss the use of neural networks for image classification and the process of passing image data and labels to the network.