G. ‘Sxplatn dhe IO exsental Nameaicel Gurrrasies un watutics wrth example - sf oon atHere are 10 essential numerical summaries in statistics with examples: 1. Mean: The average value of a dataset. + Example: The mean of the numbers 2, 4, 6, and 8 is (2+4+6+8)/ 4=5., 2. Median: The middle value when a dataset is sorted in ascending order. + Example: The median of the numbers 1, 3, 5, 7, and 9 is 5. 3. Mode: The most frequently occurring value in a dataset. * Example: The mode of the numbers 1, 2, 2, 3, 4, 4, and 5 is 2. 4. Range: The difference between the largest and smallest values ina dataset. + Example: The range of the numbers 2, 4, 6, and 8 is 8-2 = 6. 5. Variance: A measure of how spread out the data is from the mean. + Example: The variance of the numbers 2, 4, 6, and 8 is 4. 6. Standard Deviation: The square root of the variance. + Example: The standard deviation of the numbers 2, 4, 6, and 8 is 2.7. Interquartile Range (IQR): The range of the middle 50% of the data. * Example: The IQR of the numbers 1, 3, 5, 7, and 9 is 7-3 = 4. 8. Coefficient of Variation: The ratio of the standard deviation to the mean. + Example: The coefficient of variation of the numbers 2, 4, 6, and 8 is 2/5 = 0.4. 9. Skewness: A measure of the asymmetry of the data distribution. + Example: A dataset with a positive skewness has a longer tail to the right. 10. Kurtosis: A measure of the "peakedness" of the data distribution. + Example: A dataset with a high kurtosis is more peaked thana dataset with a low kurtosis.3, How when and why whould you Noamalig Blardordtize { Rescale “you dataNormalizing/Standardizing/Rescaling Data: A Guide Why? * Equitable Feature Importance: Prevents features with larger scales from dominating models. * Improved Model Performance: Many algorithms assume normally distributed data, leading to better results. * Enhanced Comparability: Makes data from different sources or scales comparable. When? * Feature Scaling: When features have vastly different ranges. + Distance-Based Algorithms: Algorithms like K-Nearest Neighbors (KNN) and K-Means rely on distance calculations. * Gradient Descent-Based Algorithms: Models like Neural Networks can converge faster with normalized data, How? 1, Min-Max Scaling: * Scales data to a specific range (e.g., O to 1). + Formula: x_scaled = (x - min(x)) / (max(x) - min(x)) 2. Z-Score Normalization (Standardization): * Scales data to have a mean of O and standard deviation of 1. * Formula: x_scaled = (x - mean(x)) / std(x)Example: Consider two features: “Age” (range: 18-80) and “Income" (range: 1000-100000). Without normalization, "Income" would have a much larger impact on model predictions. Min-Max Scaling: * Age: 18 becomes 0, 80 becomes 1. + Income: 1000 becomes 0, 100000 becomes 1, Z-Score Normalization: * Each value is transformed based on its distance from the mean and standard deviation. Choosing the Right Method: * Min-Max: When you want to preserve the original data range. + Z-Score: When you want to standardize data for algorithms that assume normal distribution. Remember: * Normalization is a crucial preprocessing step in many machine learning workflows. * Choose the appropriate method based on your data and algorithm requirements, + Always consider the implications of normalization on your Model's interpretation.4. Besutbe the various sorts of dota bateh PrOLwRINg do¢hninre a.ne ee OO oe eee bald ceprrds on aia mat mantory ord poctuuing oy [Link]- Tieathe Preuntey Proce OF lalds ody | Name Ln ah gr malive tery, pals ols ‘ ‘Shecant rg Aake'- wthe clata Coming Uyrea Hime (0d a abcoriog gon Fires coh Arvont ‘ tee ane J 4» Memon Hangin chee apoepride Size bog on tt avorlabl, manny: Agrrallur Size nigh be ; Sf, ememory GerGheitt| we heawary Goa ight: -| H) pea chee os S| Reade patmondy eceuhel bj Teocls trie after St es ben, v ate Prowsing eae te avoid muneny. ‘ Ble dette +P 3 b) Prtalletzecfion . ddd, tue deka act ints strallor badehes. rrallel prouattey » ff yo" gys-allows fe 4A Chiang “it x pecotg Y Ubfprtt Tadelus. pas CAD Spesd ep oyoall ong me expoctally 44 i cpu -Intavsire . “ pial Goel” a bea astm “fp eohich purltife Comp ger a TW do “° ep perform tasks, Abaring ucunes dala ard jrouning putt. tne, Borg tdi 'O) 0) | »62 == Ve Png D cheen protege , He Ye Leclant gue tind ele Comput He, Part talarl db Abtrihutet coggsteons :pamllh Computing anc) Long _vutaing Poti jo pore Aut Stabe da a system (on application at & 4 Speethic point g time. stats caued Abate (Kren a check point) danke failure epnnr (et obenep hi lalLow'ing Ek do atsume evilhost hase fo sdant over prom te * beg toning. Batch - Lud 4099? 09 1-. i will keep Track © mike a rvicred dunt Lte procs. de helps ab dunttigiog wrues early Inthe fey reece wf any Bede get ue Gn dutey curd pos Hy dala: ) boritentey ard Perot _ Prenttosing be bracke ben proud arc] Recreate Reseunce maregnent Fi tnditata Flat suscens ane bifng sud appictently Spo, bir 1 maniony (Ran), plo Barclesedl, pewct Comunphis, ) esting_arel vette - [Balch _eluel shsieg Raf ausing dated yaltelation ot clu bakes to Orxns LLL conedams of proun'irg dagie fx ach atte clala: | ‘pats hug tus hove octusued +%, Determine the data's dagras Bh Woalation by computing dhe kool Porson coed ont x 20 20 60 go (oo Y (00 200 200 Goo Soo -HO LY ome en 22 400 | 50d 2 fail ton logpatant the Kaa pearson cogptaent mess he Une tovrelalina between roo vortanles. ff 7anges from “2 to 4.4 value of 0 indicates no linear wrvelation - omuo - v2 (2x4) - (zur) (= y) a —_—_————— a (em) — (sai . In(e ye) Gy Avswa'- — ns. xt ye xtyy x yy 20 loo 000 400 10000 40 200 e000 1600 440000 bo 300 1g000 3600 40000 ee eae se 400 32000 6400 160000 ‘90 500 gQsv000 Em: y= Zyt = Joo 1500 650000 gyen (Zmy) (En) l2y2) oe 25 (12000) — (B00 x1s00) \| 5 (22000) - (300), J 5 (ss0000) - Usvoy a 550000 — 450000 110000 -— 40000 - 27150000 — 2260000 100000 B 190000 (41-4). (07-1) 49949-5) = 4 J. Peapectty Covnelatid -lo. Yoram owsttt Count af, (00) or 1a sladonts 0x2 shown dn dhe doltowing ° Data. 46 . AB, 1001 IOs 2143 _~ S01 10 FOURS Qo, too Desuitee ursthg the boxplot oppacach .& Box plou:- > dé % Used to weprecent Graphically , vauious Parameters at a Gane > Sn box plot, we whl! find median ,intaquartele Tonge (Tae) and Outtina . 7 A bowot is tften created ip Comput and Cont vont two (ov) move Poups - 2 biven- hoto? 4 , 98, 00, 90, WwW, 43 , 60, FO, 80, $5, 90, 100. woudrod- waite, tne dalo's ip asiending aD 60,70, 80, 95,90, 90,42, 43 595 ,9% , 100,100 Meddan /- 11 viedda. N= 12 IQR = AS% ile — asi 2 oagerg “bas x's od wm cement _ 3 element 2 as — $0 SIsputtion Lows outlion | 30 - (.exis) S 80-22.5- = 54-7 Higher oudlter a5-(.s xis) = 95 -2a.5 = 42.5 4+ To fyaw Box plot. — min Sy “ a as. Median vin Sy Qe a a @9, ith @ vo. 20,03) Gs), 48, 100,(i)