Beruflich Dokumente
Kultur Dokumente
TECHNIQUE
Department of Agroindustrial Technology, Faculty Agricultural Technology, Bogor Agricultural University, Bogor, West Java. Counselor: Dr. Eng. Taufik Djatna, S.TP, M. Si.
TRIYODA ARRAHMAN
Purposes
Related Work
Methods
Discussion Conclusion
CRM Includes Cross selling as Solution For More Effective and Efficient Marketing System
Sales Strategy is Made by Analysis of Transaction Data using Associative Rules Mining Technique
Keywords: CRM (Customer Relationship Management), Sales Strategy, Data Mining, Associative Rules Mining, Cross Selling Strategy , RFM Analysis
To Get the Support Score Of Association Rules to Know Size of Domination Level Of Itemset From All Transaction To Get the Improvement Score Of Association Rules to Know Size of Possibility Level Of Two Item Can Buy Concurrently
To Get the Confidence Score Of Association Rule To Know Size of Relation Between Two Item by Conditional
Bugher (2000)
Making Data Tables To Determine Frequency Of Each Product Item which is Sold With Another Product
Berry (2000)
Determine Cross-Selling Products That Are Suitable For One Of The Largest Banks In America which Has Millions Of Subscribers
Cashin (2003)
Adhitama (2010)
Using The Technique Of Associative Rules Mining For Produce The Sales Strategy In An Indonesia's Largest Retail Company
Start
Clustering K-Means
Threshold of item set created with trial and errors methods (Adhitama, 2010)
s( X Y )
(X Y )
N
c( X Y )
(X Y ) (X )
i( X Y )
s( X Y ) s (Y ).s ( X )
Determine Sales Strategy With Associative Rules Scores And RFM Analysis
Begin
RFM ANALYSIS
Recency: Based on The Recent Date (Updated) Customer transaction Frequency: Based on The Quantity of Products purchased by the customer Monetary: Based on The Value of Transaction made by the customer
Transaction Data
Divide Customer Into 5 Binning for analyze Recency, Frequency, and Monetary
Analyze Recency from the newest transaction (binning 5)- the oldest transaction (binning 1)
Analyze Frequency from the highest purchase transaction (binning 5) the lowest purchase transaction (binning 1)
Analyze Monetary by multiplying the quantity of purchase with the variable of each product
Assumption: (The Monetary Process is done by multiplying the Quantity of purchase with the variable of each product)
End
Begin Determine Cluster Nominal Calculate Centroid Calculate Distance Clustering Based on Minimum Distance Yes
No
Begin
System Development
System Analysis (Bottom Up), Output: - System Description - Information needs analysis - System functional requirements UML, Output: - Use Case Diagram - Activity Diagram - State chart Diagram - Class Diagram
System Implementation, Output: - Borland Delphi 7 - Power Designer 15.3 - MySQL - Ms Excel
NO
End
1. Passenger Bias (A) 2. Passenger Broad Market (B) 3. Passenger Broad Market Premium ( C ) 4. Passenger High Performance (D) 5. Passenger Ultra High Performance (E) 6. Ultra Light Truck Radial (F) 7. Light Truck Radial (G) 8. Ultra Light Truck Bias (H) 9. Light Truck Bias (I) 10. Bias Truck (J) 11. EM A 21 Bias (K) 12. EM A 3 A GDR Bias (L) 13. EM A 3 A LDR Bias (M) 14. Front Farm Bias (N) 15. Rear Farm Bias (O) 16. Ground Tire Import (P) *Data from a Tire Industry In Indonesia
bundling in connection with support value, especially combination products with a score of small support (Adhitama 2010)
2. The biggest results of multiplication between support and confidence can be used to determine strategy of sales (Novrina 2010).
3. Segmentation customers are grouped according to propensity scores, such as churn scores, cross-selling scores, and so on, which are estimated by respective classication (propensity) models (Konstantinos Tsiptsis 2009)
4. While the combination of products E and F has improvement score 1, indicating that the product E and F are positively correlated, which means that if a customer buys E product, customers also agreed to buy the F product, otherwise if the value of improvement score <1. (Adhitama 2010)
Result of Calculating of Support Score STEP 1 Product Support Product A 0.01 Product B 0.55 Product C 0.42 Product D 0.51 Product E 0.11 Product F 0.48 Product G 0.48 Product H 0.48 Product I 0.51 Product J 0.50 Product K 0.04 Product L 0.12 Product M 0.07 Product N 0.01 Product O 0.16 Product P 0.01
K-item set=2
Support 0.276
STEP 3 Confidence
0.502 0.508 0.499
Support * confidence 0.1383 0.1418 0.1817 0.1441 0.1303 0.1300 0.1371 0.1329 0.1382 0.1315 0.1301 0.1289 0.1285
0.316
0.282
0.575
0.512
0.268
0.257 0.26 0.258 0.251 0.251 0.25 0.256
0.487
0.505 0.519 0.511 0.535 0.524 0.518 0.516 0.502
0.954
1.044 1.083 1 1.104 1.026 1.015 1.034 1.005
RFM ANALYSIS
All Of Strategy : For Customer With The Highest Scores of Frequency and Recency (Binning 5) Second Strategy : For Customer With The Highest Scores Of Monetary In B and D Products
Third Strategy : For Customer With The Highest Scores Of Monetary In B, J, H, D, F, I Products
Fourth Strategy : For Customer With The Highest Scores Of Monetary In B and D Products
Sales Strategy of cross selling in this work created from support, confidence, and improvement scores.
We Found: The Biggest Size Of Dominate Level Of Rule Itemset Is When Customer Buy Product B (Passenger Broad Market Product) With 55 % Support Score
We Found: The Biggest Improvement score (1.128) is If Buy B (Passenger Broad Market) Then Buy D (Passenger High Performance)
We Found: 22 Rules Significantly For Determine Sales Strategy In Cross Selling With Calculate Support, Confidence, And Improvement Scores
Data Mining
Data Mining adalah serangkaian proses untuk menggali nilai tambah dari suatu kumpulan data berupa pengetahuan yang selama ini tidak diketahui secara manual. Data mining adalah proses untuk penggalian pola-pola dari data.
Data mining menjadi alat yang semakin penting untuk mengubah data tersebut menjadi informasi (Margaretta, 2010)
Bagian-Bagian Ban
Tread
Bagian telapak ban berfungsi untuk mlindungi ban dari benturan, tusukan obyek dari luar yang dapat merusak ban
Breaker
Bagian lapisan benang (pada ban biasa terbuat dari tekstil, sedangkan pada ban radial terbuat dari kawat yang diletakkan diantara tread dan casing
Casing
Lapisan pembentuk ban, merupakan rangka dari ban yang menampung udara bertekanan tinggi agar dapat menyangga ban
Bead
Bundelan kawat yang disatukan oleh karet yang keras, melekat pada Pelek
Jenis-Jenis Ban
Ban Bias Dibuat dari banyak lembar dengan sudut carcass cord 40 sampai 65 derajat terhadap keliling lingkaran ban Ban Radial Carcass cord membentuk sudut 90 derajat terhadap keliling lingkaran ban.
Ban Tubeless
Terdapat lapisan dari karet lembek sintesis yang disebut innerliner. Lapisan ini akan mengurung udara dan membuat ban menjadi
tubeless
Karet
Karet Alam Merupakan politerpena yang disintesis secara alami melalui polimerisasi enzimatik isopentilpirofosfat Gugus kimia: Karet Sintesis Karet Khusus yang dibuat dengan tujuan tertentu (meminimalisir kekurangankekurangan yang ada pada karet alam, Contoh: IIR (isobutene
isoprene rubber)
Trial and Errors Methods to determine thresholds of frequent itemset methods: The number of item sets (rules) should be residing until half of amount of product classification (Adhitama, 2010).
With trials and errors methods, Parameter limits (threshold) determined that min_support = 24.75% and min_confidence = 25% (result 8 the number of item sets in K=1 from 16 classification of product)
Cross Selling
Teknik menjual sesuatu barang/jasa yang berhubungan dengan suatu barang/jasa
Contoh: Seorang pembeli handycam, ditawarkan untuk membeli battery handycam, tas handycam, tiang untuk menyanggah handycam, dsb.
Metode K-Means::
PELANGGAN
P1 P2 P3 P4 P5 P6 P7 P8 P9 P10
A1
10 7 9 9 3 3 5 8 10 8
A2
26 25 25 27 24 29 24 32 24 28
A3
31 16 21 42 37 43 37 21 12 28
A4
446 266 249 213 419 238 438 348 279 342
Ya
Tidak
Selesai
MENGHITUNG CENTROID
Untuk centroid pertama, dua data pertama dianggap sebagai centroid bisa juga diacak mana yang pertama Centroid: P1 (10, 26, 31, 446) c1 (10, 26, 31, 446) P2 (7, 25, 16, 266) c2 (7, 25, 16, 266)
MENGHITUNG JARAK
Menggunakan rumus Euclidean Distance
P1 terhadap c1
P2 terhadap c1 Dan seterusnya
MENGHITUNG JARAK
P1 terhadap c2 P2 terhadap c2 Dan seterusnya
MENGHITUNG JARAK
Hasilnya disusun dalam Matriks (D0)
c1
c2
Untuk iterasi 1 Klaster1 = P1, P5, P7 Klaster2 = P2, P3, P4, P6, P8, P9, P10
MENGHITUNG JARAK
Hasilnya disusun dalam Matriks D1
c1 c2
Hasil Perhitungan
Perhitungan K=3
Dari tabel-tabel di atas, didapat F3 = { }, karena tidak ada >= sehingga F4, F5, F6 dan F7 juga merupakan himpunan kosong.
Support Calculation
transaction ID 1 2 3 Passenger Bias (A) 1 0 0 Passenger Broad Market (B) 1 1 0 Passenger Broad Market Premium (C) 0 1 0 Passenger High Performance (D) 0 0 1
4
5
1
0
1
1
1
0
0
0
Confidence Calculation
transaction ID 1 2 3 Passenger Bias (A) 1 0 0 Passenger Broad Market (B) 1 1 0 Passenger Broad Market Premium (C) 0 1 0 Passenger High Performance (D) 0 0 1
4
5
1
0
1
1
1
0
0
0
Improvement Calculation
transaction ID 1 2 3 Passenger Bias (A) 1 0 0 Passenger Broad Market (B) 1 1 0 Passenger Broad Market Premium (C) 0 1 0 Passenger High Performance (D) 0 0 1
4
5
1
0
1
1
1
0
0
0
Support A= 2/5= 0.4 Support C= 2/5= 0.4 Support AC= 1/5= 0.2 Improvement AC= 0.2/(0.4*0.4) = 1.25 (Positive Correlated)
Related Work
Proposed three price scenarios that can be applied, namely: "Together" (for example, "buy X and Y with separate $__")," prices" (e.g., "buy X for $ __, and get only the price of Y $__"), and "freebie" (for example, "buy X for $ __, and Y-free") (Harlam 1995)
Direct Marketing Association (DMA) in 1991 to determine oneto-one between product categories.
References
Adegboyega Ojo and E. Estevez (2005). Object-Oriented Analysis and Design with UML, e-Macao. Adhitama, B. (2010). "Determining the sales strategy using the association rules in the context of crm." Berry, M. J. A. a. L., G. S. (2000). Mastering Data Mining The Art and Science of Customer Relationship Management. New York, Jhon Wiley and Sons. Borland, C. (2002). Borland Delphi. version 7.0. B. 4.453. Bugher, G. (2000). "Market Basket Analysis of Sales Data for a client of Cambridge Technology Partner." Megaputer Intelligence Inc., available Cashin, J. R. (2003). Implementation of A Cross-Selling Strategy for A Large Midwestern Healthcare Equipment Company. Department of Psychology, Southern Illinois University at Carbandole. FOLDOC (2001) Unified Modeling Language. Harlam, B. A. e. a. (1995). "Impact of Bundle Type, Price Framing and Familiarity on Purchase Intention for the Bundle." Journal of Business Research, 1995, 33, pp. 57-66. Jianxin(Roger) Jiao, Y. Z., & Martin Helander (2006). "Analytical Customer Requirement Analysis Based on Data Mining." Idea Group Inc. Konstantinos Tsiptsis, A. C. (2009). Data Mining Techniques in CRM: Inside Customer Segmentation. West Sussex, Wiley. Microsoft, I. (2007). Microsoft Excel 2007. Novrina (2010) Association Rule (Algoritma a Priori). OBrien (2008). Introductory Business Information Systems Perspective Edition 7. New York, Mc Graw Hill. Oracle (2011). MySQL. Sybase, I. (2010). PowerDesigner Studio Enterprise Standalone local. 15.3.0.3248. Witten, I. H. a. F., E. (2005). Data Mining Practical Machine Learning Tools and Techniques 2nd Edition, Morgan Kaufmann Publisher.
Mendokumentasikan sistem piranti lunak UML menawarkan sebuah standar untuk merancang model sebuah sistem
The Unified Modeling Language (UML) is used to specify, visualize, modify, construct and document the artifacts of an object-oriented software-intensive system under development (FOLDOC 2001).
Activity Diagram
Activity diagram is used to describe the workflow activities in the system, in other words is how systems perform certain functions
Statechart Diagram
In general statechart diagram describes some certain class (one class can have more than one diagram statechart).
Class Diagram
Class diagram is the main diagram in object-oriented modeling. Class diagrams are used to show static structure of the system. The class is a collection of objects that have attributes and behavior (operations) which similar
Association
Dependency
Cal cul ate Support
Actor
Activity Diagram
Menggambarkan berbagai alir aktivitas dalam sistem yang sedang dirancang, bagaimana masing-masing alir berawal, decision yang mungkin terjadi, dan bagaimana mereka berakhir.
Activity Diagram
Begin Flow Swim lane Activity
End
StateChart Diagram
Menggambarkan transisi dan perubahan keadaan (dari satu state ke state lainnya) suatu objek pada sistem sebagai akibat dari stimuli yang diterima
Menggambarkan class tertentu (satu class dapat memiliki lebih dari satu statechart diagram).
Statechart Diagram
Input Username and password entry / Username and password ... [Cancel or Quit] [Submit data]
Transition
[Password and username salah]
Condition
authentification do / authentification ... [Password dan username benar]
State
input NIP Supervisor entry / NIP ... [Input NIP Success]
Class Diagram
Sebuah spesifikasi , inti dari pengembangan dan desain berorientasi objek
Gambaran keadaan (atribut/properti) suatu sistem, sekaligus menawarkan layanan untuk memanipulasi keadaan tersebut (metoda/fungsi)
Struktur dan deskripsi class, package dan objek beserta hubungan satu sama lain seperti containment, pewarisan, asosiasi, dan lain-lain.
Class Diagram
Class memiliki tiga area pokok : 1. Nama (dan stereotype) 2. Atribut 3. Metoda
Atribut dan metoda dapat memiliki salah satu sifat berikut : Private, tidak dapat dipanggil dari luar class yang bersangkutan Protected, hanya dapat dipanggil oleh class yang bersangkutan dan anak-anak yang mewarisinya Public, dapat dipanggil oleh siapa saja
Class Diagram
0..1 Rul es 0..1 rul es Perhi tungan Support 0..1 Rul es 0..* Rul es Support Rul es Item Set Juml ah transaksi i tem set j uml ah transaksi menghi tung support () ... 0..1 ni l ai support : i nt : i nt : i nt Perhi tungan Frequent Item Set Hi mpunan Item set Bi l angan i tem set data transaksi : i nt : i nt : i nt Menentukan rul es i tem set () : voi d ... 0..1 Rul es : i nt 0..1 rul es confi dence Perhi tungan Confi dence Produk + + Customer + + + Nama Customer Area : std::stri ng : std::stri ng : voi d 0..1 Rul es Improvement Perhi tungan Improvement 0..1 ni l ai i mprovement 0..1 Ni l ai i mprovement Rul es Item set Ni l ai support Base produk uni on addi ti on al produk ni l ai support base produk ni l ai support addi ti onal produk menghi tung ni l ai i mprovement () ... : i nt : i nt : i nt : i nt : i nt 0..* data produk Jeni s Produk : std::stri ng Gol ongan Produk : std::stri ng 0..1 Ni l ai Confi dence Rul es Item Set Ni l ai support Base produk uni on addi ti on al produk Ni l ai support base Produk menghi tung ni l ai confi dence () ... : i nt : i nt : i nt : i nt
0..* data customer 0..* data transaksi Fi l e T ransaksi 0..* data transaksi 0..1 Data transaksi + + + + Nama Customer Area Jeni s Produk order Gol ongan produk order Nomor T ransaksi T anggal transaksi : : : : : : std::stri ng std::stri ng std::stri ng std::stri ng i nt i nt : voi d
0..1 Rul es
0..1 Ni l ai support
Program Penentuan Strategi Penj ual an Rul es Item Set Ni l ai Support Ni l ai Confi dence Ni l ai Improvement : : : : i nt i nt i nt i nt : voi d
Menentukan Strategi Penj ual an dengan men gol ah rul es, support, confi dence, i mprovement () ... 1..* strategi penj ual an
0..* Data transaksi Marketi ng Offi cer + Nama Data T ransaksi : std::stri ng : std::stri ng : voi d : voi d
0..* strategi penj ual an Supervi sor Pemasaran 0..* Data transaksi + Username Password Data transaksi strategi penj ual an target penj ual an : : : : : std::stri ng std::stri ng std::stri ng voi d* i nt : voi d : voi d : voi d : i nt
Mendata data transaksi () Mel aporkan data transaksi () ... 0..1 Data transaksi
Association
mel akukan Log In program PSP 1 As Penggu na () mencapai target penj ual an Cross Sel l i ng () mengol ah strategi penj ual an () mengi nput data transaksi ke dal am progra m PSP1 () ... 0..* 0..1 target penj ual an
Class
T arget Penj ual an Cross Sel l i ng Juml ah nomi nal penj ual an
: i nt : i nt : i nt : i nt
Eval uasi target penj ual an Cross Sel l i ng () mengawasi j uml ah nomi nal penj ual an () ...
0..* 0..1 j uml ah nomi nal penj ual an j uml ah nomi nal penj ual an
: voi d* : i nt : voi d
menerapkan strategi penj ual an cross sel l i ng dal am penj ual an () ...