Page 190 - AI Ver 3.0 class 10_Flipbook
P. 190

As a result, the app clusters the photos into categories like "vacation," "family gatherings," or "pets," without
              needing any input or labels from the user.
              This example shows how unsupervised learning works by discovering hidden patterns in unlabelled data, just like
              the child in the swimming pool explores and learns independently.
              Let us take an example of the Unsupervised Learning - Fraud Detection:
              A bank processes a large number of transactions daily, maintaining a database with details like:

              •  Transaction amount
              •  Transaction location
              •  Time of transaction
              •  Account activity patterns
              The  goal  is  to  identify  potentially  fraudulent  transactions.  However,  the  transactions  are  not  pre-labelled  as
              "fraudulent" or "non-fraudulent."
              An Unsupervised learning algorithm is applied to analyse the data and group transactions based on patterns.
              The algorithm automatically identifies unusual behaviours, such as:
                 • Unusually large transactions.

                 • Purchases from unusual locations.
                 • Multiple transactions within a short time frame.
              The model forms two main clusters:
                 • Cluster 1: Regular transactions that follow typical patterns.

                 • Cluster 2: Suspicious transactions that deviate significantly from normal behaviour.
              Based on these clusters, the bank can flag transactions in Cluster 2 for further investigation, without needing prior
              labels for fraudulent activity.
                                                                                   Regular Transactions
                             Total No.   Total No. of                         80   Suspicious Transactions
                Transaction
                            of Regular   Suspicious
                   ID                                                         60
                           Transactions  Transactions      Unsupervised
                                                            Learning          Transaction Frequency  40
                                                             Model            20


                                                                               0
                                                                                     50      100    150     200     250
                                                                                           Transaction Amount ($)
              This example illustrates how unsupervised learning helps uncover hidden patterns in unlabelled data, enabling
              more effective decision-making.
                                                                                        21 st  Century   #Media Literacy
                                                                                            Skills
                          Video Session

                     Watch the video on "Neural Networks" at the given link:
                     https://www.youtube.com/watch?v=jmmW0F0biz0 or scan the QR code and answer
                     the following question:

                     What are the five things that you should know about neural networks?






                    188     Touchpad Artificial Intelligence (Ver. 3.0)-X
   185   186   187   188   189   190   191   192   193   194   195